Here’s Why.

Within the span of a few weeks, two major studies on online search behavior circulated through the American AI Search & SEO community. Two rigorous studies, two documented methodologies, two recognized authors in their respective fields. And conclusions that appear to contradict each other.

• Rand Fishkin — founder of SparkToro, published on March 2, 2026, an analysis conducted with Datos on desktop search behavior in the United States.

  → AI tools account for 3.2% of searches. Google still dominates at 74%.

Conclusion: The hype around ChatGPT is disconnected from behavioral reality.

• Ethan Smith — CEO of Graphite, an agency specializing in SEO and AEO (he is, incidentally, actively pushing to popularize the term AEO over GEO to describe AI Search) — published a study drawing on Similarweb data.

  → AI now accounts for 56% of global search volume and 34% in the United States. ChatGPT alone represents 20% of global search traffic.

Conclusion: AI is underestimated, not overestimated.

Both studies went viral.

They were read, shared, and hotly debated. Rand even commented on Ethan’s LinkedIn post, calling it “nonsense.”

Before diving into the details of each, let’s frame the two core questions at stake:

• The more obvious one: What is the actual size of AI Search?

• The underlying one: How do you select the right data to inform your AI Search strategy — and, by extension, make sound marketing budget allocation decisions in 2026?

Over the past two years, we have witnessed an explosion in the number of studies published on AI Search — from agencies, SaaS platforms, independent researchers, and media outlets alike.

The volume has become so significant that it is increasingly difficult to separate genuine data from marketing demonstration. Because behind every study, there is an author. And behind every author, there is an industry, a positioning, a business model.

This is not a value judgment. It is simply a reality we need to factor into how we read and interpret this data. Many of these studies are not worth retaining or analyzing in depth.

These two are.

I’ll break them down for you — decoding their findings, examining their limitations, and unpacking their disagreement — and, as always, I’ll share my 2 cents on the broader trend I draw from both the substance and the form of these two studies. Because taken together, in any case, they open the door to an increasingly fascinating debate.

Let’s get into it.

AI Search Is Overestimated. Here’s the data.

👉 Sparktoro study published on March 2, 2026..

For 25 years, Google was the gateway to the internet. Searching the web was synonymous with searching on Google.

Over the past two and a half years, Google has lost its monopoly on internet search. Other tools have entered the equation.

This does not mean the end of Google — it means the addition of new ways to search and new platforms on which to search. The second layering on top of the first, not replacing it.

The central question the latest SparkToro and Datos study set out to answer was one that matters to any marketer:

→ Where does Search actually happen on the internet in 2026?

From the outset, Rand Fishkin raises a framing question that cuts to the heart of the matter: if we call Claude and ChatGPT search tools, why wouldn’t we also call Instagram, YouTube, or Amazon search tools?

This is precisely the famous platform shift that Sundar Pichai never tires of referencing — and the inescapable constraint it places on us as marketers to perform the “platform dance,” given that every platform now operates its own proprietary algorithm.

Rand’s answer is unambiguous: search happens everywhere.

Calculating total search volume on the web requires including every single place where humans type queries to find information, a product, a service, or an answer.

Methodology:

The methodology is rooted in the annual analysis conducted with Datos, covering the 250 most visited sites in the United States and Europe.

From those 250, Rand Fishkin selected 41 sites, divided into five categories — traditional search, commerce, social, AI, and other verticals.

Two immediate observations worth flagging upfront:

• 14 sites alone reach more than 20% of Americans,

• and the analysis is conducted exclusively on the desktop.

The study aimed to answer four key questions:

1. Where does search happen on the internet?

2. How has the share of search evolved throughout 2025?

3. What is the difference between a “Visitor” and a “Searcher”?

4. How has the number of searches per user evolved in 2025?

Question 1 — Where Does Search Happen on the Internet?

The study’s baseline finding is unambiguous: ChatGPT is far smaller than commonly assumed.

Google remains responsible for nearly 74% of all desktop searches across the 41 domains analyzed in the United States, in Q4 2025.

• The full breakdown of desktop search behavior is as follows:

• Traditional search engines capture 80% of searches.

• Commerce sites — Amazon, Walmart, Booking, Airbnb, eBay — capture 10%.

• Social networks come in third at 5.5%.

• And AI tools currently represent 3.2%.

The numbers tell a story that is radically different from the surrounding AI hype.

The report highlights three major trends in traffic behavior:

1. First trend: the decline of Wikipedia. This is directly linked to the drop in traffic sent by Google. A complementary observation worth noting: Wikipedia is nevertheless massively cited within LLMs. ChatGPT in particular is the platform that scrapes Wikipedia most heavily. This is a perfect illustration of the current moment — less direct traffic, but more citations within AI tools.

2. Second trend: the rise of Instagram’s desktop traffic. Historically a mobile application with very limited desktop functionality, Instagram has firmly established itself on computers as well.

3. Third trend: ChatGPT’s growth followed by stagnation. What everyone had already identified.

4. But it is the fourth observation that is the most significant — and the least discussed. Even before ChatGPT, other domains are generating more desktop search activity:

   • Amazon — unsurprising.

   • Bing — more surprising.

   • YouTube — owned by Google.

These three platforms receive more desktop searches than ChatGPT, today, in 2025. And marketers have invested infinitely less in them than they have in LLM optimization over the past three years.

The collective obsession since 2022: being referenced in ChatGPT.

The result: Amazon, Bing, YouTube, and Google have been neglected in favor of a platform that, in real behavioral volume, remains marginal on desktop.

For Europe, the trends are broadly identical, with three nuances. Google is slightly more dominant there — approximately 5 percentage points higher than in the United States. Amazon, YouTube, and Bing hold somewhat smaller shares. And ChatGPT is used marginally more — about 1 percentage point higher — consistent with the strong adoption of AI tools across Europe.

Question 2 — How Has the Share of Search Evolved Throughout 2025?

The fundamental shift in search behavior is not a migration toward AI. It is a transformation in how we search. We no longer perform a search the way we used to. We inject prompts.

To analyze trends beyond Google — too dominant to allow a clear visualization of the other platforms — the axis was adjusted to start at 70% market share. Two distinct trends then emerged clearly.

1. First trend: the 34 sites outside the top 7 increased their share of search. This is counterintuitive. The conventional assumption is that the very largest sites become ever more dominant over time. Here, the smaller platforms gained ground. This is the inverse of the scenario most commonly anticipated.

2. Second trend: Google declined slightly, in favor of Amazon, Bing, and YouTube — all of which grew considerably. ChatGPT, meanwhile, also declined — and not by a negligible margin.

In Europe, the picture is slightly different: the top 7 is more stable, Google declined less, ChatGPT declined less, and the most striking growth belongs to Amazon.

Question 3 — What Is the Difference Between a Visitor and a Searcher?

This is perhaps the most important distinction in the entire study — and the one most frequently omitted from audience analyses.

The difference between a visit and a search.

A person who visits a site is not necessarily a person who performs a search on that site.

Many users arrive on a platform without ever typing a single query: the scrolling behavior — you land on LinkedIn or Instagram, you consume, you do not search.

• Platforms where more than 70% of visitors perform a search include Google, DuckDuckGo, Bing, Craigslist, and Amazon. Logical — these are tools whose primary purpose is search.

• The rate drops to approximately 50% for eBay and for ChatGPT.

Why is ChatGPT at 50%? When you land on ChatGPT, nothing prompts you. You don’t see other users’ conversations. You are alone facing the chatbot — and if your intent isn’t already clearly formed before you arrive, you leave without having searched for anything at all.

Why is eBay also at 50%? Category browsing, auctions, sales tracking — everything is displayed immediately upon arrival. The user doesn’t need to search to find what they came to see.

Direct conclusion: using visits as a proxy for platform usage is a biased metric — particularly for AI tools. A visit and a prompt are two fundamentally different things. Analyses built on traffic data rather than actual prompts are actively misleading.

Question 4 — How Has the Number of Searches per User Evolved in 2025?

Rand Fishkin divided the 41 sites into three groups based on the volume of searches or prompts per user per month.

Group 1 — More than 10 searches or prompts per month

ChatGPT is the only AI tool in this group — and it dominates clearly within it. Google, DuckDuckGo, and Amazon also feature here. Only Bing experienced a slight decline over the course of 2025. The overall trend for this group is stable.

Group 2 — Between 5 and 10 searches or prompts per month

Two trends stand out sharply.

• Pinterest is showing remarkable growth — already flagged in the Datos and SparkToro State of Search report. A platform not to be underestimated.

• A steep decline for eBay, by contrast.

Group 3 — Fewer than 5 searches or prompts per month

LinkedIn appears here, which surprises Rand Fishkin himself. He had expected it to be a more popular search destination. The explanation comes down to a single sentence: Google is more efficient for searching within LinkedIn and Reddit than their own internal search engines. Users don’t search on LinkedIn — they search on Google, and LinkedIn results surface from there.

A notable and unexpected data point: the growth of Threads. From roughly two searches per user to approximately four over the course of 2025 — a much more intensive usage pattern. In January 2026, Threads surpassed Twitter’s daily active user count on mobile.

👀 My Two Cents

1. — Search happens everywhere humans have intent. Going online is synonymous with “searching.” The marketer’s priority — since the very beginning of marketing — has always been to place their message where their potential buyers are looking. This is not new. What is new is that we had forgotten it.

2. — This is the conclusion that changes everything. Search is not a behavior. It’s a channel. Which means the real job is not to optimize for Google or for ChatGPT. It’s to identify where your audience searches — and to be there. Fishkin calls this Search Everywhere Optimization.

3. — We really need to stop with the tired (and frankly unfounded) “ChatGPT has killed / will kill Google” narrative. But the number does need refining. Google is still too often cited at 90% market share. That figure is wrong — or at the very least, incomplete. The study places it closer to 70%. And probably nearer to 65% in reality, since the 41 sites analyzed exclude all long-tail editorial properties that would mechanically bring that figure down further. In Europe, Google is more dominant than in the United States — around 80% — but still far from the 95% that traditional methodologies continue to report.

4. — Most AI searches are already happening on Google — not on ChatGPT. This is the least discussed point in the entire study. Even adding up every single ChatGPT, Claude, DeepSeek, and other LLM prompt combined, Google crushes them all. Why? Because 16% of its results now display AI Overviews. Google is not losing the AI battle. It is already, by a wide margin, the number one AI search engine in the United States and Europe. With one exception: France, where AI Overviews have not yet been deployed — which considerably changes the equation for the French-speaking market.

5. — The study assumes equivalence across all types of search queries. This is an acknowledged limitation, not an oversight. But it is precisely this point that Smith will attack — and which we unpack in the next section.

AI Search Is Underestimated. Here’s the data.

👉 Graphite’s Study Mars 2026

Graphite’s study on the size of AI Search draws on Similar web data, enriched with prompt intent data from an OpenAI study covering approximately one million messages.

It covers five major LLMs — ChatGPT, Gemini, Perplexity, Grok, and Claude — and six major search engines: Google, Bing, Yahoo, DuckDuckGo, Yandex, and Baidu. Other discovery surfaces — Amazon, social networks — fall outside the scope of this study.

The Opening Finding: AI Is Underestimated, Not Overestimated

Ethan Smith begins from an initial hypothesis that is the inverse of his conclusion. He originally believed that AI usage was overestimated. After analyzing the data, he concludes the opposite: previous estimates were significantly undervalued — sometimes by a factor of 4 to 5.

Why this gap?

Because the majority of prior analyses only accounted for desktop web data — entirely omitting mobile.

Yet 83% of AI sessions worldwide take place via mobile applications. In the United States, that proportion is 75%. Comparing Google’s web visits with ChatGPT’s web visits while ignoring mobile usage is the equivalent of measuring half the phenomenon — at best.

The Key Numbers

Monthly AI-related sessions now represent 56% of global search volume and 34% in the United States. These figures are 4 to 5 times higher than previous estimates, which were based solely on web data.

In absolute volume, AI totals 45 billion monthly sessions worldwide, of which 5.4 billion are in the United States.

But Ethan Smith makes an important distinction. AI usage related to search — meaning question-type prompts, which he terms “Asking” prompts — represents 28% of global search volume and 17% in the United States. Because not all prompts are the equivalent of search queries. According to OpenAI’s classification, only 52% of prompts are of the Asking type.

[INSERT VISUAL: Global vs. US AI Sessions — comparison with search volume — Similarweb data]

The Breakdown of Prompts by Intent

The OpenAI study on which Smith draws classifies prompts into three categories:

• Asking prompts — seeking information or advice — represent 51.6% of usage.

• Doing prompts — requests to execute tasks — account for 34.6%.

• Expressing prompts — personal expression — make up 13.8%.

Only Asking prompts are relevant when comparing AI to traditional search. And even within this category, not all of them are interchangeable with a classic search query — some involve complex conversations or requests for advice that would never have been entered into a search engine in the first place. Smith acknowledges this directly in the study.

AI and Search: A Positive-Sum Game?

This is one of the most important conclusions in the study — and one of the rare points on which Fishkin and Smith implicitly agree.

Since the launch of ChatGPT, the overall use of search — traditional engines plus AI combined — has increased by 26% globally and 16% in the United States between Q1 2023 and Q4 2025.

Search is not dying. It is expanding. AI is adding to search rather than immediately replacing it.

This follows the same logic as the arrival of mobile applications in 2008. At the time, Wired magazine ran a cover headline in September 2010: The Web Is Dead. Long Live the Internet. Mobile apps did indeed reach massive adoption. But the web never disappeared — the market simply expanded. Smith draws this parallel explicitly.

ChatGPT (Would) Dominate the AI Market — and (Would Be) Gaining Ground on Google

ChatGPT accounts for 89% of AI sessions worldwide and 86% in the United States. This represents near-total domination of the LLM market.

Against traditional search engines, the dynamic is real.

Google’s share of traffic would have fallen from 89% in 2023 to 71% by the end of 2025. ChatGPT would now represent 20% of global search-related traffic and 12% in the United States.

AI Is More Significant Outside the United States

Global usage is more than 7 times higher than US usage alone. This figure deserves serious attention: the bulk of AI growth is happening in the rest of the world, not in the United States. At the global level, AI sessions have stabilized since July 2025.

In the United States, however, usage continues to climb, with a 300% increase in December 2025 compared to December 2024.

The Methodological Limitations Smith Himself Acknowledges

The author is transparent about the limitations of his work. The correlation between Similarweb data and proprietary data — Google Analytics, Search Console — is very high for web data, with a median correlation of 0.86. However, Smith explicitly acknowledges that he is unable to independently verify the mobile data. This is precisely the point on which Rand Fishkin will intervene.

The Size of AI Search: Impossible to Measure. And Impossible to Ignore.

Following the publication of the Graphite study, Rand Fishkin responded publicly on LinkedIn:

“I don’t believe this data. It doesn’t match anything others have observed, the linked Google Doc cites neither an original source nor a methodology, and the 83% mobile figure seems dubious at best (and the only source appears to be Graphite, the agency itself). Sorry, but I have to say this looks like complete nonsense to me.”

Ethan Smith responds:

“The article I published clearly defines all metrics, includes links to the raw data as well as to the Similarweb sources. All data comes from Similarweb, with the exception of queries classified by user intent, which come from OpenAI. Graphite is the source of none of the data. I am simply describing Similarweb’s data. I find it strange that you continue to characterize our research as dubious — or even fraudulent — without formulating any specific criticism.”

The End of the Saga — With a Further Response, dated today, March 17.

Ethan Smith subsequently published a detailed response to these criticisms. In it, he offers two clarifications that merit close attention.

1. Rand Fishkin had initially used ChatGPT’s American MAUs (Monthly Active Users) to challenge global figures — a scoping error he himself acknowledged and corrected. This does not change the substance of the disagreement, but it does temper the force of the critique.

2. Ethan Smith acknowledges that he did not fully account for Google’s mobile app usage in his calculations. This is potentially significant — if Google’s mobile usage is underestimated, the share attributed to AI in his figures could mechanically be overstated. He indicates he is working on an updated version.

What Is the Debate Really About?

The tension between these two studies is real — but misframed.

Those two studies don’t contradict each other. They are just not measuring the same thing. The problem is that they are presented — and perceived — as directly comparable.

Desktop vs. Mobile

The SparkToro study analyzes desktop only. Explicitly. Its scope: 41 sites, desktop behavior, Q4 2025. Result: AI represents 3.2%.

The Graphite study incorporates both web and mobile applications. And this is where the numbers explode: 83% of AI usage takes place via mobile. By excluding mobile, you exclude the majority of real-world LLM usage. Graphite’s result with mobile included: 56% of global search volume.

The 83% Mobile Figure

Rand Fishkin contends that the source of the 83% mobile usage figure is Graphite itself. In other words: the author of the study would be his own source on the single most decisive point of his entire argument.

Smith’s response: all data comes from Similarweb, not from Graphite. Graphite is merely describing Similarweb’s data.

What is unresolved?

Similarweb’s mobile data cannot be independently verified. Something Ethan Smith himself acknowledges in the methodology section of his study. The correlation with Google Analytics and Search Console applies only to web data — not to mobile sessions.

What Counts as a Search

• The SparkToro study only counts behaviors where a user actually types a query on a platform. It distinguishes between visit and prompt, and shows that only 50% of ChatGPT visitors actually perform a search.

• The Graphite study starts from total AI sessions, then applies a filter: Asking prompts represent 52% of all prompts according to OpenAI. Even these prompts are not all interchangeable with a classic search query — Smith acknowledges this — but he uses them as a proxy.

The Scope of Platforms Analyzed

• The SparkToro study analyzes 41 sites including Amazon, YouTube, Bing, Reddit, Pinterest, and Threads. It encompasses the full spectrum of search behavior on the web. Result: AI represents 3.2% of a very broad market.

• The Graphite study compares only 5 LLMs against 6 search engines. It explicitly excludes Amazon, social networks, and other discovery surfaces. Because its reference market is narrower, the share attributed to AI is mechanically higher.

The Points of Convergence Between the Two Studies (Unspoken)

Both studies agree on one central point that nobody cites enough. The core of the demonstration.

The SparkToro study shows that Google remains dominant but that alternative platforms are gaining ground. The Graphite study quantifies overall search volume growth at +26% across both engines and AI.

This is not AI v/s Traditional Search. It’s :

What is the real size of AI in this new landscape, how do we measure it — and above all, what is the strategic importance of knowing this number when designing your own content strategy?

What’s the real size of AI Search in 2025–2026?

Understand the subtext:

Does it actually make sense to invest your marketing budget in this AI Search that everyone is talking about — but that very few people truly know how to evaluate and measure?

And then, the question within the question:

How do you sort through the data — identifying what is reliable enough to inform your content strategy and guide where you invest your €€€/$$$ — versus what is noise?

Both questions are worth asking.

Why?

→ From a brand perspective, publishing original research that leverages your proprietary data is one of the most effective content formats for standing out from the “AI slop soup” that AI itself despises — and it is the dream content format for being sourced by AI, positioning your brand as trustworthy, with sufficient authority and legitimacy for AI to recommend you. This is what I call creating “WOW content.”

→ From a marketer’s perspective, knowing how to read these studies — and sort the genuinely objective content that truly informs a trend worth factoring into your content strategy (and therefore your budgets) from the irrelevant — is a real and pressing skill.

Over the past two years, I have never witnessed such a proliferation of studies on AI Search.

Semrush, Conductor, SparkToro, Graphite, Webflow, Brightedge, BrightLocal — and many more. Dozens of tech brands and marketing agencies are publishing reports, analyses, and benchmarks. Especially since AI Search and GEO are already considered multi-billion dollar industries after just a few years of existence.

I have been tracking AI Search’s pulse every single day for nearly three years, and the volume of my Notion curation is growing at a staggering pace and speed. The rhythm is accelerating. Genuinely.

And with it, the risk of confusing data with demonstration.

Because behind every study, there is an author. And behind every author, there is a positioning, an industry, a business model. This is not an accusation. It is a reality that every informed reader must factor into how they consume this content.

Two studies, two visions of AI Search?

• The SparkToro study builds a precise argument: AI Search is not synonymous with exclusive optimization for LLMs. It is synonymous with Search Everywhere Optimization. Its central conclusion is both powerful and correct: search is not a behavior, it’s a channel. What this means in practical terms is that before investing in optimizing your presence within AI tools, you need to analyze where your audience actually spends its time. And that audience does not necessarily spend its time in LLMs — it may spend its time on YouTube, on Amazon, on Pinterest, on Threads. Rand Fishkin founded SparkToro, an audience intelligence tool designed precisely to answer this question: where does your audience live online?

• The Graphite study moves in the opposite direction.

The agency has positioned itself as one of the most advanced players in optimizing brand presence within AI agents and answer engines. Its expertise, its business model, its reputation are all built on the conviction that AI Search is the immediate future of digital marketing. AEO is the solution.

AI Search = One More Marketing Channel

Regardless of which figure you adopt — 3.2% or 56% depending on the methodology — one conclusion transcends the debate:

Search is not a behavior. It’s a channel. And AI Search is simply one more — with its own rules, its own audiences, its own visibility mechanics.

What this means in practical terms for marketers is a hierarchy of decisions — not a checklist of platforms to tick off.

Andy Crestodina, founder of Orbit Media, describes precisely what has changed:

His reading of the transformation is surgical.

• Informational traffic has collapsed: Google’s AI Overviews answer directly. LLMs do too. For a large portion of informational queries, the answer is already there, on the page, with no click required. Discovery content marketing — the idea that a blog post generates massive organic discovery — is over, or nearly so.

• Content remains indispensable: for emails, for social networks, for sales teams, for existence and for being cited. But it no longer generates the mass traffic we once knew it for.

What changes everything is the shift from visibility to recommendation.

→ Appearing in an AI response for a query with commercial intent is one thing.

→ Being recommended is another. And that is what matters now.

Because the prospect does a large part of their deliberation inside the model — within the generated responses — before ever landing on a website.

What this produces is measurable: conversion rates from ChatGPT are meaningfully higher than those from Google. The prospect arrives warm, qualified, often already decided.

Regardless of the channel — Google, AI Overviews, ChatGPT, Claude, Copilot, Gemini — the job remains the same: be present, feed it, be recommended.

From Data to Action: 4 Strategic Decisions for Integrating AI Search

Being present everywhere your audience searches no longer serves only to be found by humans. It also serves to be ingested by AI models.

LLMs don’t only scrape websites. They scrape Reddit, YouTube, LinkedIn, specialized forums, niche communities, public databases.

Wherever content exists and carries authority, AI absorbs it. This means a brand present on YouTube with authoritative videos in its sector is not only optimizing for YouTube. It is also feeding the models that will recommend it tomorrow in their responses.

The existing internet — with all its platforms, communities, and formats — is becoming the training and citation ground for AI models.

Multi-platform presence is therefore no longer purely a reach strategy. It is an AI algorithm-feeding strategy.

The question is no longer just “where does my audience search?” but “where does my audience search AND where will AI models draw from to recommend me?”

In many cases, the answer is the same. Which makes the multi-platform presence strategy doubly profitable — immediately for humans, structurally for the models.

#1 — Map Before You Optimize

Audience knowledge is the single foundational building block that will transcend every era and every major disruption of the web and marketing at large. On the web or off it, before the internet or after, with Google or with LLMs — marketing and sales are simply impossible without it. This is not a conviction. It is a constant. It existed before the internet. It will exist after generative AI. It will survive every technological disruption because it is not a technique. It is a principle.

Before launching an AI Search strategy or chasing citations in LLMs, analyze your audience: does my specific audience actually live there? Not the global audience. Not the aggregated figures that vary by a factor of five depending on the methodology. The audience of your sector, your market, your product. If it is on YouTube, on Amazon, on specialized forums — that is where you need to be visible first. And that is also where AI models will go to draw from.

#2 — Do Not Neglect Under-Invested Platforms

The desktop data makes it clear: Amazon, Bing, and YouTube receive more searches than ChatGPT.

These are three surfaces where marketing investment remains far below the actual audience volume. For a brand whose audience spends time on these platforms, this is a direct opportunity — often far less competitive than LLM optimization. And as a bonus: these are also sources that AI models cite massively.

#3 — Treat Google as the First AI Search Engine

16% of Google results display AI Overviews. Google is already, by a wide margin, the largest AI search tool in volume. Optimizing for AI Overviews — and therefore for AI search — means optimizing for Google first. Your SEO fundamentals remain relevant, even if they are no longer sufficient on their own.

With one exception: France, where AI Overviews have not yet been deployed — which changes the equation considerably for the French-speaking market (unfortunately).

#4 — Integrate LLMs If Your Audience Is There — and Measure

ChatGPT traffic converts better than Google traffic. The prospect arrives warm, qualified, often already decided — research done inside the model before they ever land on your site.

If your audience is in LLMs, being recommended is not optional. It’s a priority.

But being recommended NOT EQUAL as being cited. The real stakes are no longer visibility — they’re active recommendation. And that changes everything about the content you need to create.

Want to go deeper? Join Andy Crestodina and me in Chicago on April 22nd. 4 hours to understand exactly how AI decides who gets recommended — and how to make sure it’s you 😉.

For the first time since the advent of modern search engines, the direct correlation between a brand’s visibility and traffic to its website is broken.

• SEO World: Visibility = Traffic. Being seen = being clicked.

• New GEO World: These two metrics are decoupling.

This is the Visibility Without Traffic paradox: “the great decoupling.”

Your brand can be:

• Massively mentioned in AI-generated responses (high visibility)

• Synthesized and exposed to millions of users via ChatGPT, Perplexity, or Google AI Overviews

• Recommended as the reference for “the best payroll software” or “the best credit card”

• Cited as a source of truth in thousands of responses

... all while recording a historic drop in web traffic.

Your content serves to build the answer, satisfies the user, but generates no visits.

And you’re not even aware!

The Invisible Funnel: Micro-transactions out of sight

An example?

A user asks ChatGPT: “Compare HubSpot and Salesforce for an SMB.” The AI performs the analysis, compares features, evaluates pricing, recommends HubSpot, and the user closes the tab satisfied.

• HubSpot won the interaction

• Salesforce lost

• Traffic to the sites? Zero for both

Your Analytics shows “no impact,” but reality shows a sale potentially lost or won. These are millions of micro-influence transactions now happening “out of sight,” in what we call the Invisible Funnel – a conversion tunnel that traditional attribution models are unable to measure.

The brand “won” the interaction, influence was exerted, brand preference was built, but the website remains deserted.

In other words, if you’re only measuring traffic, you’re steering your strategy in the dark.

The Halo Effect Without Visits

This is the Billboard Effect.

Millions of people see your brand in the AI’s response. They don’t click, but they imprint your brand in their minds. Users treat AI responses as factual summaries. They scan the content, memorize the cited brand, but don’t click immediately.

The value has shifted: it no longer resides in the site visit (immediate conversion), but in the influence exerted upstream, at the moment of reading the response.

If you cut the budget because “clicks” are dropping, you disappear from their minds and lose the consideration battle.

The Quality Paradox

The Pew Research Center reports a striking figure: while users say they’re satisfied with AI summaries, only 1% of them click on the citation links included in these summaries.

What a paradox? Traffic drops, but the quality of that traffic increases.

Data shows that visitors coming from conversational experiences (LLMs) can have conversion rates up to 6 times higher than traditional Search.

The user who arrives on your site after an AI interaction has already:

• Asked their follow-up questions

• Refined their need

• Validated your credibility via the engine’s synthesis

• Compared alternatives

They’re no longer coming to search, they’re coming to act.

The clicks that survive this filtering are no longer curiosity or discovery clicks, but high-stakes transaction or verification clicks (particularly in YMYL domains - Your Money Your Life).

Volume drops, but the intent behind each remaining click is exponentially stronger.

The trap

If you’re steering your strategy on traffic volume, you’re blind.

Your Analytics shows “no impact” while millions of micro-influence transactions are happening out of sight.

You risk cutting the budgets that build your authority while your competitors silently dominate the citation space..

The Window of Action Is Closing

This rupture isn’t theoretical. It’s happening now.

→ 12-24 month horizon: Gartner predicts a 25% drop in traditional search volume by 2026. If your business model relies on organic traffic to fuel your pipeline, you have 12 to 24 months to react before irreversible impact.

→ Explosive adoption: 52% of American adults already use LLMs. To contextualize this speed: generative AI is reaching critical mass faster than mobile. The smartphone took 10 years. ChatGPT reached 100 million users in 2 months.

→ B2B: 95% of buyers plan to use generative AI in their purchasing journey this year. This is no longer experimentation. It’s default behavior.

Three Business Risk Scenarios

Your exposure depends on your model. No one is immune.

SCENARIO 1: The “Aggregator” Model — Existential Risk

• You are: Comparison sites, directories, “Top 10” blogs, informational portals living off discovery traffic.

• Your historical value: Gathering scattered information and presenting it in an organized way.

• The problem: AI does exactly that, instantly and better. Your added value is commoditized. Users no longer come to your site for a list. They ask the AI: “Compare the 5 best CRMs for an SMB” and get a synthesized answer in 10 seconds.

• The impact: Your model relies on volume and ad impressions. A 25-40% traffic drop (already observable) collapses your revenue proportionally. You don’t have 24 months. You have 6-12 months to pivot.

SCENARIO 2: The “Lead Generation” Model — Severe Risk

• Who are you? B2B SaaS, professional services, marketplaces dependent on organic traffic to fuel the funnel.

• Your historical value: Capturing attention at the top of the funnel with informational content, then converting to leads.

• The problem: The Top of Funnel is evaporating. Informational queries (”What is a CRM?”) are satisfied directly in the AI interface. Users no longer enter your funnel. They bypass it.

• The impact: If organic represents 40-60% of your leads, your CAC will explode. You’ll have to compensate for lost free traffic with expensive Paid. Or not compensate and watch your pipeline dry up.

• The attribution trap: AI influences your prospects upstream (invisible). They arrive as “Direct” (visible). You attribute success to the wrong channel. You risk cutting budgets that work — expert content, digital PR — because they no longer generate measurable direct clicks.

SCENARIO 3: The “Transactional” Model — Moderate but Real Risk

• Who are you? E-commerce, established brands, retailers selling directly.

• Your historical value: The final destination to buy.

• The problem: AI becomes the personal shopping assistant. It pre-qualifies, compares, guides the decision before arrival on your site. If your brand isn’t cited in this initial conversation, you don’t exist in the “consideration set.”

• The impact: You don’t necessarily lose total volume. You lose market share to brands that AI systematically recommends. If your competitor dominates AI citations, they capture a disproportionate share of qualified intent.

• The Long Tail risk: You historically captured traffic on thousands of niche queries. AI aggregates these niches and recommends 2-3 dominant brands. If you’re not in that top 3, you disappear.

Facing these clearly identified risks today, a classic mistake:

Treating GEO as just another acquisition channel.

GEO isn’t a tactic. It’s a redesign of your marketing infrastructure.

The question to ask yourself to address one of these three risks is no longer “if” but “when” you’ll be impacted.

And above all: do you have a plan before your competitors dominate the citation space you’ve spent years building?

🟣 I’m Karine. I help you adapt your content strategy to AI Search and GEO through a masterclass and a community focused on AI marketing. The next session of my GEO masterclass in english will be in May. Want to join?

Join the waitlist

GEO (”Generative Engine Optimization”) is not a simple “evolution” of SEO or an algorithm update—it’s a paradigm shift.

An earthquake.

Our entire relationship with information (access, consumption, engagement) is being transformed.

Google is no longer the only gateway to the internet. There’s ChatGPT now.

Beyond our keywords, the economic fundamentals of our online businesses have changed.

All of this in less than two years.

New web. New rules of the game.

And a lot of questions that sting...

Join the waitlist

Moving from SEO to GEO: OK but... 🤕

🔴 The Traffic Paradox (1% Volume, 100% Intent)

Traffic from AI stagnates below 1%. AI keeps the user for information and only sends them to you for the final action.

👉 Question: How will you fuel your sales funnel and create brand preference if AI confiscates the entire education and discovery phase?

🔴 The “Invisible Sale” (Decision Inside, Conversion Outside)

AI acts as a salesperson who advises and convinces before the click. There’s now a third party between your future customer and you. And it can recommend you—or not...

👉 Question: What happens to your brand if the purchase decision is made on ChatGPT or Perplexity before the customer even arrives at your site?

🔴 The Measurement Crash (Analytics Blindness)

AI operates in a “black box,” outside your Google Search Console. Clicks disappear, but influence remains.

👉 Question: How do you justify your budgets and drive your performance when your dashboards have become myopic? Now we measure “citation share,” not clicks. Are you ready?

🔴 The Obsolescence of “Blog Content” (The 20/80 Rule)

On Google, people search for information. On ChatGPT, they want to do something. 80% of prompts aim to write, code, analyze, create. Only 20% seek pure information.

👉 Question: Today you write to please AI. So it cites you. If it doesn’t cite your content, it’s invisible. Is your “Content” budget burning for nothing?

🔴 Financial Risk

If “free” organic visibility disappears in zero-click responses, dependence on Paid risks exploding to maintain volumes.

👉 Question: Can your business model withstand the shock if your Customer Acquisition Cost doubles? Have you thought about another way to allocate your budgets?

🔴 The Trust Deficit (UGC Risk)

AI engines “snub” brand sites in favor of Reddit, press, and third-party reviews deemed more neutral.

👉 Question: Do you have an off-site influence strategy, or are you leaving your reputation (and therefore your AI visibility) in the hands of forums you don’t control?

🔴 The Death of the Keyword (Semantic vs Lexical)

An average prompt on ChatGPT is 60 words. Users no longer type keywords, they chat. We’re moving from keyword to intent and context.

👉 Question: Are your teams still optimizing for keywords that nobody types, or are they building the semantic universe necessary to be “understood” by the machine?

🔴 The “One-Shot” Urgency (52% Churn)

On Google’s new AI Mode, 52% of users test once and don’t come back the next day.

👉 Question: Is your content tailored properly to be sorted and then cited by AI? Or is your super detailed content lost in the Algo + LLM mix?

🔴 Siloed Organization

GEO requires perfect structured data (Tech), strong brand authority (PR/Comms), and useful content (Marketing/Product).

👉 Question: Who owns the topic internally? If the answer is “just SEO,” we need to talk.

Join the waitlist

I know...

That’s a lot of questions.

I’ve given more than twenty conferences and workshops in recent months to executive committees, CMOs, solopreneurs whose business model relies entirely on their Google ranking.

97% of them had no answers to these questions.

Yet everyone has definitely felt their traffic collapse.

I can confirm, you’re not alone: 600 million organic visits lost in less than 12 months according to fresh Similarweb data on the US market.

A bit like we’re all trying to switch to electric, while keeping our combustion engine.

I’ll stop there 🤐.

Questions without answers have never really been my thing.

And as my boss used to tell me when I was a lawyer: “If you walk through my office door with a problem, make sure you also arrive with solutions.”

📈 Give Me 1.5 Hours, We’ll Build Your GEO Plan

That’s why I compiled my 2 years of AI Search “24/7 hands in the grease” to design a GEO Masterclass.

Not an SEO training, not an AI tutorial.

A strategic framework to transform a technical topic into an action plan and help you avoid the classic mistake: throwing all your dollars at tactics before identifying the new budget components.

For whom?

For executives, CMOs, and leaders who need to understand before changing everything. Who need to (re)plan before spending.

👉👉 1.5 hours to design your GEO strategy, adapt your content strategy, and reallocate your SEO budgets in the AI Search era.

From SEO to GEO, yes.

But strategy before tactics. Always.

Join the waitlist

December 2025. Third year of generative AI. And we’re finishing with a bang.

Similarweb just dropped a gem: “2025 Generative AI Landscape: The State of Gen AI”.

I don’t do this often. But this one, I really dissected it. And not with AI. Old school (like when I used to review findings): printed, highlighter in hand.

Why? In the flood of “stats & data” reports that flooded last year — often cobbled together in 10 minutes via “deep research” by tech startups with a panel of 90 clients — Similarweb is, in the US, a technological giant.

This is not an anecdotal player: 5,308 clients worldwide, over 100 million websites and 4 million apps tracked.

When Similarweb publishes a report, these aren’t just numbers: these are insights into the real global Web. Robust, representative, massive data.

And in this case, this latest year-end report is overflowing with killer data.

Numbers that your GEO strategy — where the goal is to be visible in AIs, not just “ranked well on Google” — should exploit.

But also because it confirms what I’ve been saying for 15 years:

📌 Content strategy is the center of gravity of every business. It deserves a seat at the executive table — not just in the SEO/Content team.

The finding is clear Generative AI is the new connective tissue of the web.

In less than three years, it went from “futuristic gadget” status to invisible, omnipresent infrastructure.

It didn’t replace tools. It slipped between them. It opened a new gateway to the Internet — which now coexists with the first.

Silent? Yes. Brutal? Absolutely.

For brands, it changes everything: The rules of online visibility are completely rewritten.

In this note 👉 I’ve summarized the “big numbers & big trends” for you.

In this post 👇👇, we go into detail.

Side Note: If one of your business objectives for 2026 is to exist in AI responses, sign up for my GEO masterclass. 1h30 to understand the impact of AI Search and GEO on your content strategy — and know how and where to reallocate your marketing, content and SEO budgets.

Join the waitlist

1️⃣ AI — Our Best Assistant, Everywhere, All the Time

Generative AI adoption is growing fast on web (+76%)... but exploding on mobile (+319%).

👀 My take:

AI is no longer just an occasional tool (web). It’s becoming a personal pocket assistant (app). AI is inserting itself into daily routines (personal, professional, fun). Mobile transforms AI into an everyday companion.

💡 Integrate into your GEO strategy: Think global. Think user on the move. Think immediate answers.

2️⃣ Stop Optimizing for Clicks. Optimize for the Answer.

In three years, ChatGPT has claimed its place in the global Top 5 websites. Ahead of Amazon, just after Instagram 😳.

👀 My take:

Google was an intermediary between the user vs the rest of the web. AI becomes a destination. Users no longer seek the source (Google) → they seek the final solution (answer, text, code). But AI isn’t a destination like other platforms on the list: it’s a destination that gives us an answer, does stuff for us and talks with us. We’ve moved from information seeking and time spent “scrolling” to seeking immediate solutions and useful time where we exchange, are challenged, create, are more productive.

💡 Integrate into your GEO strategy: Optimize for the answer. Not for the click. Which moves us from keywords (to get the click) to intent (to provide the best possible answer).

Unlike passive attention platforms (YouTube, Instagram), ChatGPT established itself through active intent: each interaction aims for a specific goal — getting an answer, creating, solving.

AI isn’t a search engine, it’s an answer engine. It selects the best information, not the best page.

➡️ New standard: In this model, precision trumps filler: if your content rambles, AI ignores it. It chooses sources that provide the most direct solution to user needs. Your content must be useful, structured and actionable, ready to become the exact piece of answer that AI inserts into its conversation.

3️⃣ Web Micro-fragmentation and the Era of “Search Everywhere”

The web is crystallizing. Google and YouTube remain volume leaders but are hitting their glass ceiling. Facebook continues its decline. Instagram holds steady. ChatGPT is the statistical anomaly.

👀 My take:

Global traffic is no longer dispersing, it’s locking within 5 or 6 mega-platforms. The open web gives way to closed ecosystems (Walled Gardens) that capture the entire user journey. The user journey fragments according to intent:

• I want to learn/create → ChatGPT (LLM)

• I want to buy → Amazon (Marketplace)

• I want to watch → YouTube/TikTok (Video)

• I want to verify → Google (Search)

It’s also the end of “passive scrolling”. The nature of what we do on a “Top Site” has mutated. We’re moving from the era of Consumption (Search & Social) to that of Action (Creation & Conversation).

💡 Integrate into your GEO strategy: It’s the death of monolithic (Google-centric) SEO and the new imperative of Search Everywhere Optimization. Each platform has its own sovereign algorithm (Social Graph, Search Graph, or AI Model). You can no longer just optimize your site for Google. You must adapt your content to the “language” of each algorithm. You also can no longer seek to bring everyone to your site, but go yourself to those platforms where your audience is to be the essential answer there. One strategy per algorithm.

4️⃣ It’s Not Google vs ChatGPT — It’s Both

People haven’t replaced Google. They’ve added ChatGPT.

• 95% of ChatGPT users also use Google

• 441 million shared users

👀 My take: Despite what all those who wanted to see (finally) the end of Google’s monopoly era hoped, Google hasn’t ceded market share to ChatGPT. Between them, they’ve expanded the market. And each took its place. Usage is cumulative, not substitutive.

💡 Integrate into your GEO strategy: SEO!! Your GEO strategy includes SEO. Period. Google remains the anchor point for verification and transactional navigation (so you continue your SEO); ChatGPT captures the upstream research phase (so you must also invest in GEO)

5️⃣ Focus on Where Adoption is Massive: ChatGPT

Claude, Perplexity, DeepSeek are rising... But ChatGPT dominates very largely (by the way, not a word, not a comma about Mistral...). It seems we’ve traded one monopoly for another: ChatGPT is 80%+ of generative AI usage. It’s the “Google of AI”, the default standard. At least until now.

👀 My take:

We’re witnessing live specialization of uses. ChatGPT’s hegemony hides qualitative fragmentation. We’re moving from a single tool to a “Toolbox”. Users are starting to choose their AI based on the task:

• Generalist / Daily → ChatGPT

• Precise research / Sources → Perplexity

• Writing / Nuance → Claude

• Google ecosystem → Gemini

💡 Integrate into your GEO strategy: You’ll quickly realize that “being visible in AIs” isn’t the right approach when you start getting your hands dirty. Each LLM has its own “black box” to dissect to understand how to be visible “at their place”. Since we’re not approaching this topic at a total loss, the priority is therefore first to be visible in ChatGPT. Not visible on ChatGPT = not visible in AI. However, your very next focus should clearly be AI Mode. 2026 will tell us if Google’s firepower will shift users from ChatGPT to its AI mode (Google’s direct response to ChatGPT).

6️⃣ AI Mode Reaches 100M Visits — But Only 10% Return

In countries where AI Mode was deployed (I know, not yet in France 🤬) its adoption was lightning-fast:

AI Mode became the fastest Generative AI platform to reach 100 million visits in the United States (at the same time, given Google’s firepower it’s not exactly a surprise).

Except that... Lots of tests. But not yet integration into habits. 50% don’t come back.

A notable segment of about 10% returns regularly (5+ days), signaling the beginning of a loyal user base.

Personally, I use AI Mode every day - and frankly I love it. I’ve done in-depth cross-research with creation tasks and between us, integration with all my Google workspace, NotebookLLM, my Gems, My Drive etc. make it gold in terms of usage.

👀 My take:

Habits die hard. And ChatGPT having taken everyone by surprise when 3.5 was released took the lead on “Chat” usage with us. When we have a question or anything, we spontaneously go to ChatGPT. Google has incredibly caught up in 3 years for sure. But bringing a company the size of Google up to speed happened gradually. Google SGE in 2023, Google AI Mode in 2024 and AI Mode + hundreds of AI Apps fully integrated into our workspace in 2025. The friction phase AI Mode vs ChatGPT is normal. I personally think it won’t last.

For you in France, without AI overviews or AI Mode as I write these lines, it’s not easy to determine an approach. But what emerges here is basically...

💡 Integrate into your GEO strategy: Seize the “Zero-Click” opportunity in AI Overviews. Since people rarely go directly to AI Mode, they start with a Google search that increasingly leads to an AI Overview (see the screenshot I just took below 👇👇), which generally offers you, at the end of the response, to switch to AI Mode. So to anticipate presence in AI Mode, you must aim to already appear in AI Overviews.

7️⃣ AI is Aging (in a Good Way)

AI is no longer almost exclusively Gen Z’s toy. It’s leaving the early adopters’ circle to become mainstream usage.

Dilution of Young People: The share of 18-34 year-olds declines (from 61% to 53%), not because they’re leaving, but because others are arriving.

Rise of Seniors: 45+ year-olds now weigh 30% of the audience.

Consequence (in the form of SUPER opportunity): Purchasing power (often higher among 45+) has entered the conversation. AI is becoming a mature commercial channel.

👀 My take:

THE key concept to remember from this data: the principle of “decoupled decision” (decision inside, conversion outside). Users decide which product to buy during their conversation with AI. They only go to the website to execute the transaction. AI plays the role of the in-store salesperson who advises and convinces. The website is now just the cash register. If you haven’t convinced in the LLM, the customer will never arrive on your site.

💡 Integrate into your GEO strategy: You can no longer treat AI as a simple performance channel - meaning judging it on direct traffic or immediate click. You must measure both brand and performance. And logically, rethink attribution and KPIs accordingly. In particular you must measure your “Citation Share” (how many times does AI recommend me?) and your “Brand Sentiment” (how does AI talk about me?). In this new ecosystem, influence precedes the click: so optimize your content to feed AI’s sales pitch (comparisons, expert reviews, clear benefits).

8️⃣ Our Browsers Are Becoming “Intelligent”

Call them Atlas (OpenAI), Comet (Perplexity). The year 2025 marks the end of the passive browser (Chrome/Safari).

These browsers are no longer empty windows to display sites. They’re intelligent overlays capable of “reading” the screen with you. They don’t just respond, they act. They click, fill carts and send emails in place of humans.

👀 My take:

We’re moving from “Read-Only” to “Do-It-For-Me”. From the Information era to the Action era. Users no longer just want to find the product or info, they want the browser to do the work. The browser mutates: it goes from a consultation tool to an Autonomous Agent. The friction between “having the idea” and “executing the task” disappears.

💡 Integrate into your GEO strategy: “Agent-Ready” Optimization. It’s a new frontier for GEO (and here we go with yet another acronym: AIO (Agentic Interface Optimization 😩). If your site is too complex, stuffed with pop-ups or technically unreadable for a robot, the Agent (Atlas/Comet) won’t be able to execute the task for the user. It will go to a “compatible” competitor. To avoid this, you must expose your functionalities (purchase, booking, registration) so that an AI can activate them without friction. Your site must become a visual API for Agents. Structure your data, simplify your forms. You’re no longer designing just for the human eye, but for AI’s hand.

9️⃣ The “Search Intent” Fracture

The numbers shatter the idea that ChatGPT is just a “new Google”.

Google = The Librarian: It serves to retrieve info (informational) or go to an address (navigational).

ChatGPT = The Workshop: Only 20% (1 in 5) of prompts serve to search for info. The remaining 80% serve to create, analyze, rewrite or solve.

Transactional: Google keeps the advantage on immediate purchase (8.7% vs 2.1%).

👀 My take:

We’re moving from retrieval to creation: Users change mental mode according to the tool. On Google, they’re in passive mode (”Find me that.”); on ChatGPT, they’re in active mode (”Do that for me.”). Except that... If users aren’t seeking info but action, then the very definition of “Search” is obsolete. It’s really the shift from the knowledge economy (clicking on links) to the competence economy (solving problems).

💡 Integrate into your GEO strategy: If 80% of prompts aren’t information questions, your classic blog articles (”What is X?”) are useless on AI. So your content must focus on providing tools, not just facts. It must help AI work for the user. If your content helps AI write the email, structure the project or debug the code, you’ll be cited. If you only explain the concept, you’ll be ignored in favor of AI itself. Content that works: Templates, frameworks, checklists, pre-written prompts, data analyses. Clearly, short keyword SEO is in a state of clinical death. Enter the conversational cluster.

1️⃣0️⃣ Search Becomes a Conversation

One prompt on ChatGPT = approximately 60 words

One search on Google = approximately 3.4 words.

The conclusion is quite clear: Search becomes dialogue. And what connects the two is the user adding context to their search (personalization, precision, preferences, explanations, etc.).

The finding is the same on Google’s AI Mode, even if usage is still slow (see 👆👆).

👀 My take:

Search evolves from a transactional act (finding a link) to a conversational act that leads to concrete action (solving a problem, creating something). Users who start an AI session expect iteration and a response that takes their specific constraints into account from the first exchange, until final satisfaction (which could be: I found the right brand for buying my next running shoes 😉.

💡 Integrate into your GEO strategy: Google already gave us direction with RankBrain then Hummingbird: end of post catalogs built only on keywords aligned haphazardly. Since 2015, sites built in topic clusters (in French “semantic cocoons”) are favored (the E-E-A-T era): a site structured in thematic clusters, not as a “keyword library”, are most likely to demonstrate strong editorial authority. With AI Mode and its Query Fan-Out logic, Google takes us to the next level: It’s no longer enough to cover the entire lexical field of your topic. You must also cover all indirect logical connections around the problem. Your content must be an exploitable, structured, contextualized, actionable answer.

1️⃣1️⃣ AI Shuns Brand Sites — It Prefers Authority & UGC

When ChatGPT and AI Mode need to retrieve information to build their responses, they don’t draw uniformly from all types of web sources. A clear hierarchy emerges, and it doesn’t favor content that brands publish on their own sites.

Sources favored by ChatGPT and AI Mode:

✅ Authority content: news sites, media, specialized press

✅ User-generated content: reviews, forums, community platforms

❌ Brand/institutional site content: not popular.

👀 My take:

AI seeks social validation and supposed objectivity. Information present only on your site is considered marketing; the same information validated by media or discussed on a forum becomes a “truth” cited by AI. It’s the comeback of social proof as a technical ranking factor - and PR...

💡 Integrate into your GEO strategy - You must massively invest in spaces you don’t control (not yet): Digital PR and Community Management on Reddit/forums become GEO pillars. Goal: get talked about elsewhere to be visible in AI’s response. But... as I repeat in my workshops and masterclass: Saying “in 2026, we’re opening a subreddit” isn’t a strategy. It’s a tactic. However: Deciding to invest in a community space on a third-party platform (WhatsApp, Reddit, Circle, Slack, or even a forum at home) 👉 that’s ✅

1️⃣2️⃣ AI-Referred Traffic Already Stagnating: <1% of Organic Traffic

After an explosion phase of referral traffic from generative AI platforms in 2024, end-2025 data reveals early (and worrying) stagnation for those hoping to see AI become a major acquisition channel. Expert analyses like Kevin Indig confirm that AI’s model resembles nothing like Google’s.

• LLM traffic: <1% of total organic traffic

• Less than 1% of users click in AI Overviews (Pew Research)

• Early stagnation end 2025

👀 My take:

The “Zero-Click Search” era has reached maturity. ChatGPT isn’t a traffic engine like Google. It’s an influence platform like TikTok or TV. The goal: brand impression, not visit. Kevin Indig confirms: “ChatGPT is closer to TikTok than Google.”

💡 Integrate into your GEO strategy - Radically change your KPIs: abandon referral traffic volume. And especially, don’t evaluate your GEO success by referral traffic volume in GA, you risk being disappointed. Adopt a new KPI: Share of Voice. Brand impact exists before the click. AI must speak positively about you and cite your arguments, even without a visit. Design your content to deliver value directly in the response (Brand Awareness), not as a teaser to force the click.

1️⃣3️⃣ Google AI Mode vs Traditional Google Search

There it is, we expected it: Google’s AI mode generates less traffic than Traditional Search (no kidding!). Direct comparison between click-through rates of Google Search and AI Mode reveals even an abysmal gap.

• Google: 17–19% redirection

• AI Mode: ~2%

This is easily explained: Google Search and Google AI Mode have fundamentally different missions - and claim it! Google was designed as an intelligent web directory, while AI Mode aims to synthesize and resolve directly in the interface.

That said, signals suggest that with habit and adding real-time data, clicks will eventually rise.

AI Mode doesn’t seek to copy Google’s traffic model. It reinvents the journey between question and answer, keeping users in the interface, at the heart of the AI experience.

👀 My take:

Google is a routing tool (finding where info is). AI is a synthesis tool (understanding info) and action aid. You need to change mindset: consider “Zero-Click” no longer as a bug or inevitability, consider it as one of the main features. User satisfaction comes precisely from not needing to click. As for the remaining 2% of clicks, if by chance they’re maintained in the new journey Google is reinventing

💡 Integrate into your GEO strategy - You don’t want to depend ad vitam on traffic Google will decide (or not) to send you - we’re talking about 2% huh... - when it determines the new “mix of Search and AI Mode” model. Start from the following assumption: Don’t seek to get users out of AI (it’s counter-intuitive for them), rather seek to influence what they read inside. Your content strategy must aim to indelibly associate your brand with your industry, and your niche. Go all-in on everything that can develop your brand awareness, starting with personal branding - that’s what leaves the best impression.

1️⃣4️⃣ Low Traffic, But Ultra-Qualified

If traffic volume from AI remains marginal, qualitative analysis of this traffic reveals - for now - a real difference. Visitors who arrive from an AI recommendation behave radically differently from those who arrive via Google, with engagement metrics that defy all comparison (data-backed): clearer intent and higher engagement.

• Google traffic: high volume, low engagement

• AI traffic: low volume, but time spent ↑, pages viewed ↑, conversion rate ↑↑

👀 My take:

Google traffic is “mass traffic” often superficial. AI traffic is “niche traffic” ultra-qualified. Visitors from AI explode engagement metrics (time spent, pages viewed, conversion) compared to those from Google. AI acts as a pre-qualification filter, basically your best salesperson. Users have already roughed out the topic with the chatbot. If they finally decide to click, it’s no longer to discover, but to act or verify. The click out of AI isn’t exploratory, it’s decisional.

💡 Integrate into your GEO strategy - Adjust your landing pages. If a user arrives from AI, they already know a lot about your brand. Don’t welcome them with generalities. Strengthen your “Services”, “About us” pages. Offer immediate added value (deep expertise, tool, exclusive offer) that complements AI’s summary, without repeating it. AI just made a super pass; content on your page must score the goal ⚽🥅. An AI visitor may be worth 10 Google Search visitors. Without precise data, calculate your ROI on final conversion, not session volume.

1️⃣5️⃣ ChatGPT Becomes Super App: PayPal, Spotify, Everything Happens Inside

ChatGPT is mutating: all recent announcements of third-party app integrations directly into ChatGPT prove it. Their goal goes well beyond being just a conversation tool (which everyone understood they’d struggle to monetize and burn a lot of cash scraping even more data). They want to become a unified interface where users can accomplish concrete tasks (do EVERYTHING there, especially buy...) without ever leaving the environment.

• Major integrations: PayPal (planned 2026), Spotify

• Search explosion following October 28 announcement

• Users can buy, create playlists... without leaving the interface

👀 My take:

Already consider that ChatGPT is no longer a chatbot, it’s an action platform. It’s the beginning of the “Agents” era. AI no longer just serves to retrieve information, it serves to execute the task. Friction related to changing applications disappears. ChatGPT aims to become a “Super App” where the entire customer journey, from discovery to purchase, happens internally. Everything that can be done without leaving the interface will be.

💡 Integrate into your GEO strategy - Even if we still enjoy some traffic, we need to face it: that’s not where the future is being played out anymore. Medium term, even visibility - which is today the first GEO objective - will no longer be sufficient. You must be “actionable”. If your competitor is natively integrated (via plugin or API) and you’re not, they’ll sweep up the transaction even if your content is better. You’ll need to think “API First”. For e-merchants, the ultimate strategy will no longer be just having text indexed by AI, but having a product catalog connectable so AI can place orders for users.

1️⃣6️⃣ From “Ranking” to “Selection”: AI Recommends

There was “be visible in SERPs”. Now there’s “be visible in AI”. It’s the same word: visible. But the entire mechanism behind it is fundamentally (and definitively) transformed.

The fundamental difference between traditional SEO and optimization for AI (GEO) lies in the very nature of what the algorithm does. Google displays a list of identical results for everyone; AI builds a personalized and argued response.

• SEO = static results identical for everyone

• GEO = probabilistic unique response for each user

• AI actively guides decision by explaining “how” and “why”

👀 My take:

You must move from Appearance to Selection. In SEO, winning = being in position 1. In GEO, winning = being the argued recommendation by AI. AI acts like a store salesperson who must understand your product to credibly recommend it.

💡 Integrate into your GEO strategy - Your content must justify the recommendation; it must feed AI’s “reasoning” by combining psychology (understanding needs) + logic (choice criteria). Don’t just sell the product, sell why it’s the best choice for such profile. Come on, I’ll give you your new “mantra”: don’t optimize to be found, optimize to be chosen.

1️⃣7️⃣ The 6 New AI Visibility KPIs

In the generative AI era, your classic SEO dashboards become blind. Traditional metrics (average position, CTR, session volume) don’t capture what’s really happening in AI conversations.

Like it or not, your new measurement framework draws more from Public Relations and outdoor advertising campaign measurement indicators (never thought I’d write that one day talking about my job 😱) than from classic performance marketing.

The 6 GEO KPIs:

• Visibility: % of your brand mentions on relevant prompts

• Share of Voice: your visibility vs competitors (benchmark)

• Domain Influence: frequency your site is cited as source

• Top Competitors: who wins most mentions on your topics

• Sentiment: tone of mentions (positive, neutral, negative)

• Brand Strength: your sentiment score vs competitors

👀 My take We’re abandoning purely technical SEO metrics (Position, CTR, Volume) to adopt Public Relations metrics. In GEO, reputation is superior to ranking. AI (also) issues judgment. Being visible isn’t enough; you must be perceived positively (Sentiment) and cited as authority (Influence).

💡 Integrate into your GEO strategy - It’s the “Halo” effect of GEO: success isn’t measured by immediate click, but by repercussion on your overall brand. If AI recommends you, users trust you and type your name into Google (Branded Search ↑). Consider AI traffic as “Bottom of Funnel” traffic: low volume, maximum conversion. Work all content that associates your top-level expertise with your brand: patents, hyper-technical product characteristics. Everything that allows users to associate your name with being a super expert in your subject.

1️⃣8️⃣ Retail vs Banking: Visibility Varies Radically by Sector

Comparative analysis of AI visibility across different sectors reveals major disparities. In some domains, a few players dominate and crush the conversation; in others, the market remains open. Structural differences that obviously require different strategies according to industries.

US numbers:

• Retail: Target (67%) and Walmart (59%) = massive duopoly, +50% visibility

• Banking: fragmented market, even leaders struggle to reach 10%

• Travel: Airbnb, Booking.com, Expedia dominate.

👀 My take:

There’s no universal “good visibility rate” (alas). AI structure replicates market structure: extreme concentration for e-commerce, fragmentation for services.

💡 Integrate into your GEO strategy: Your competitive benchmark must be strictly sectoral. For Retail (concentrated markets): don’t attack giants head-on on generic queries. Target hyper-specific niches and long tail. For Banking (fragmented markets): the field is open, an aggressive GEO strategy can double your competitors quickly. One thing is sure however: volatility is VERY high. One thing to change then, compared to the SEO era: your visibility audit is no longer an annual/semi-annual document, it’s become a weekly dashboard.

1️⃣9️⃣ Deep Content Wins

AI digs deep: 33% of citations come from level 3 or more

Technical analysis of URLs cited by ChatGPT reveals counter-intuitive behavior: AI largely ignores main site pages (homepage, category pages) to fetch content much more buried in the architecture.

• 33.3% of AI sources = depth level 3 (site.com/folder/subfolder/page)

• Only 11.6% from level 1 (main pages)

• AI (almost) completely ignores your homepage

👀 My take:

Specificity beats global authority. It’s not enough to be “known”, you must show and explain what you’re known for. Unlike a human who enters through homepage, AI enters through the kitchen window (specific buried page). The deeper a page, the more precise it is. AI seeks the exact answer, not the store window.

💡 Integrate into your GEO strategy - The real end of “Thin Content”: generic “fluff & blah blah” pages precisely generated by AI... are invisible to AI. Develop a “Deep Content” strategy: multiply very niche pages (technical FAQ, ultra-detailed product sheets). Don’t target “Sports clothing” but “Best anti-UV running t-shirt for sensitive skin”. That’s where AI looks for its source. And the best way to do this is create AI generation topic clusters.

2️⃣0️⃣ Reddit (54%) Crushes Brand Sites (9%)

Analysis of sources cited by AI on product queries (example: “Running Gear”) shows brutal hierarchy. Forums and specialized media dominate massively, while official brand sites (even stars) are relegated to oblivion. It’s a total inversion of traditional SEO logic. You must change your approach fundamentally, in form and content.

Example stats on “Running Gear”:

• Reddit: 54% of citations (community)

• Runners World: 35% (PR)

• Nike.com: 9% 😳 (Brand)

👀 My take:

“Don’t be the source, be IN the source.” This must become your mantra (I know... another one). AI grants more credit to third-party validation (what others say about you) than your own discourse. Technical credibility shifts from your domain to those of curators (Press) and users (Reddit).

💡 Integrate into your GEO strategy - Adopt a “Barnacle GEO” strategy (cling to the big ship). Part of your SEO budget must shift toward Digital PR and Influencer Marketing. For example, identify the 3 media and 2 subreddits that dominate your sector. Your goal isn’t to surpass or copy them, but to be cited in their content. Since you can’t beat the platform, infiltrate it. Partnership, partnership, partnership!!

2️⃣1️⃣ GEO’s New “Long Tail”

Ethan Smith (CEO of Graphite) formalizes an essential distinction: GEO/AEO rules aren’t the same depending on whether you target broad, popular queries (”Head”) or very specific, long tail questions (”Tail”). Each query type, according to its popularity, requires a different content strategy.

How it works in GEO:

On big queries (Head: “Best CRM tool”), being first on your site no longer suffices. AI seeks cross-validation (YouTube, Reddit, press).

On precise queries (Long Tail: “Best CRM tool for plumber with billing”), AI seeks the exact answer, often absent from classic SEO because deemed “too niche”.

👀 My take:

AI functions on consensus for general, on precision for specific. To be recommended as leader, AI must see your name everywhere. It doesn’t believe what you say, it believes what everyone says about you. For niches, it thirsts for details that traditional SEO ignores (low search volume).

💡 Integrate into your GEO strategy - Your content strategy must split in two: Off-Site Strategy (awareness) 👉 saturate third-party platforms to create consensus. On-Site Strategy (relevance) 👉 create ultra-specific pages (Programmatic SEO, detailed FAQ) for complex questions. That’s where AI will come pick. Two major observations here but completely counter-intuitive for SEO professionals: 1) it’s time to work on keywords weakly or never queried - those can make the difference. 2) it’s also time to ask yourself THE critical question: Who owns the subject internally? If the answer is “just SEO”, you need to follow my GEO masterclass.

2️⃣2️⃣ SEO Feeds GEO. Don’t Abandon It.

A recent event (September 2025) brutally recalled the fragility of generative AI’s ecosystem: when Google modified access rules to search results for LLMs, limiting their vision to the top 10 results, impact was immediate and measurable on citations (even the stock market didn’t like it).

• September: Google limits LLM access to results beyond Top 10

• Immediate result: drastic drop in Reddit citations in ChatGPT

• Financial impact: Reddit stock decline a few days later

👀 My take:

SEO conditions GEO. In any case, it remains a solid foundation of visibility. AI doesn’t scan the entire web in real-time. It depends on what Google/Bing show it first. Clearly then, strong visibility in Google Search still reinforces visibility in AI. Yes, AI visibility is ultra-volatile - because it’s especially very dependent on data access rules. Logically the “Leader Premium” (even if it doesn’t guarantee winning the battle) reinforces: traditional SEO remains a good entry ticket. Don’t bet everything on a single third-party platform. An algorithm change can cut this faucet overnight. Action: establish weekly monitoring. In GEO, an acquired position is never permanent.

💡 Integrate into your GEO strategy - SEO 🤗.

Sources:

• https://www.similarweb.com/corp/2025-generative-ai-landscape/

• Similarweb Customers and Revenue Statistics

• https://dcfmodeling.com/blogs/history/smwb-history-mission-ownership?

My goal during this keynote was to provide you with answers to the following questions:

• What is the value of content in the AI Search era?

• What new objectives should we assign to content?

• What is the new playing field for content?

• What are the fundamentals of an AI-optimized content strategy?

• What are the new content formats imported by the generative AI wave?

🟣 What is the Value of Content in the AI Search Era?

In 1996, B. Gates’s essay published on Microsoft’s blog compared the first Internet content model to broadcasting: companies pushed information to a passive audience, much like TV viewers. B. Gates even wrote:

“Content is where I expect much of the real money will be made on the Internet, just as it was in broadcasting.”

“Content” was then text and software created by professionals and distributed via a one-way channel. The historic “Content is King” truly summarized the first 25 years of the internet as we knew it... at least until November 2022.

ChatGPT 3.5 in November 2022: The Shock

The initial reaction from most content creators (including myself): panic mode.

“If an AI can produce in seconds what took me several days, in 3 years I’ll be unemployed.”

This concern quickly gave way to apocalyptic discourse. “SEO is dead,” “TOFU content is dead,” “Google is finished,” “websites are has-been.”

This type of reaction is a classic pattern in the face of every technological disruption. Fear breeds generalized anxiety where everyone convinces themselves that everything is doomed.

The Chat Shock, the Reality of Content

For years, content was considered an asset.

👀 Its sole purpose: to be distributed via the Internet, which played the role of universal distribution channel. This channel, open 24/7 and with low production costs, enabled a real revolution by democratizing creation.

In this paradigm, the end user had a passive role: they consumed information, read, and clicked to navigate. This was the first generation of the web.

Today, AI has only accelerated a transformation already underway: the shift to a new ecosystem.

The internet is no longer a simple channel, but a set of platforms.

The main objective of these platforms? Keep the user as long as possible (you’ve probably already noticed the crappy reach close to -30% of your Facebook post where you inserted an external link to your blog article 😩).

Content has changed function.

It’s no longer an asset meant to be passively consumed, but a tool designed to generate interaction.

👀 We’ve moved from a passive consumption model to a paradigm where content serves as a pretext for dialogue, engagement—a like, a comment, a share. Content has become the trigger for conversation.

From Open Web to “Walled Gardens”

The content landscape has become much more complex, interactive, and user-centric.

Content is now a dialogue that includes user-generated creations and interactive formats: blogs, short videos, podcasts, quizzes, lives, AR/VR experiences.

A content creator is now a vast and diverse group of individuals—”creators” or “influencers.” AI tools assist and optimize creation workflows. The audience consists of active participants who share, comment, create, and influence a brand’s narrative through user-generated content.

AI-Assisted Content or AI-Created Content?

AI has obviously accelerated this transformation, we’ll see how. But to move beyond bar talk where everything would be either catastrophic or miraculous, we need to look closely at the numbers.

From the content perspective, we’re not being replaced by AI. And it’s not happening tomorrow.

However, AI has inserted itself into our creation process. Before, we started from a blank page: find the theme, structure, write. It took considerable time. Today, we use AI not to generate content, but to help us create our content.

Two recent studies perfectly illustrate this evolution:

• The blogging study conducted by Orbit Media’s Andy Crestodina, which analyzes how bloggers now use content

• Ahrefs’ study, “State of AI and Content Marketing,” which surveys content professionals on their AI usage

Concretely, what do we use AI for?

• To brainstorm, instead of organizing endless meetings

• To correct, proofread, translate

• To get ideas. I use it a lot to identify what’s missing in my content, what I forgot

• Or to rephrase when it’s poorly expressed, to condense ten sentences into one

• And almost never to create complete content

👀 Today, except for beginners, only 4% of content marketers use AI 100% to create content from scratch. AI has become an assistant, not a writer.

🟣 The New Objectives Assigned to Content

The SEO and content ecosystem is currently bathed in apocalyptic discourse: end of the web, end of SEO, end of traffic, end of clicks, death of media. A real festival of pessimism. Total depression. Heavy... Every week, following AI news for my marketing newsletter, I tell myself we could legitimately want to drop everything and change careers (I think about it sometimes, I admit).

BUT...

We must differentiate between the photo and the film.

The photo of today? It’s true. Traffic drops, sometimes up to 66%. Clicks collapse. Media are in a very difficult situation. These observations are real and verifiable. Does that mean we’re all doomed? No.

👀 The photo is what it is. But the film tells another story. Behind this crisis, we all—marketers and entrepreneurs—have a major opportunity.

Traffic Drop and Zero-Click: A Reality

The Pew Research Center study is striking (I had spotted it this summer): zero-click explodes (in green on their graph) while traffic collapses (in orange).

When this study came out this summer, Google hastened to downplay it, saying everything was fine, that AI Overviews were great, and that sites continued to prosper. I had even done a retrospective on Google’s somewhat ridiculous storytelling zigzag.

Except I have dozens of case studies that prove otherwise. B2B, B2C, small bloggers, big brands: traffic plummets everywhere. It’s factual.

Zero-Click: Nothing New Under the Sun

Rand Fishkin, brilliant and true US marketing guru who popularized SEO in the early Google era and founder of SparkToro, invented the expression “zero-click search.” This situation where the user on Google no longer clicks on one of the ten blue links on the search results page to find their result by surfing the web.

With the arrival of ChatGPT and AI Search solutions like Perplexity, the situation has worsened.

Except in reality, this problem has been latent for 10 years. Honestly, saying Google sends less and less traffic to websites is a bit like saying climate change is worsening year after year. It’s just become an even more visible storm due to ChatGPT’s success and conversational AI.

The numbers are accurate: between 25 and 27% of users no longer click on Google because they already have their answer.

But on the other hand... Google has never hidden its game.

Since 1995 and the injection of backend AI into its algorithm to better detect search intent, Google, initially a search engine, has transformed into a platform that offers content and wants to keep its users. Like all other platforms.

👀 This is called the new SEO: where Search Engine Optimization is replaced by Search Everywhere Optimization (I even made it a section in my newsletter).

abonné

Yes, we’re alarmed about the end of the click, but we were already getting there progressively. AI just accelerated the movement. And yes, it’s also true: media are taking the first hit.

AI and Media Drama

In the United States, complaints are pouring in. In France and Europe too, lawsuits are piling up. Media outlets are protesting every day.

Honestly? I sometimes have trouble crying with them.

Google with its AI Overviews, ChatGPT, Perplexity give informational answers. Naturally, that kills media whose job is precisely to give information. It’s logical.

Except, deep down, what are we surprised about?

Google is an advertising agency. That, from day 1, has never changed...

Their brilliant innovation from 25 years ago—which consisted of inventing natural referencing to make paid referencing even more credible—aimed to educate us to click on links as the preferred navigation mode on this giant new medium that was the Internet at the time.

No one should be surprised: their goal was never to direct traffic to natural results. Google’s ultimate business? Make us click on ad links.

It’s the story of the glass half empty or half full:

We benefited from natural traffic for 25 years. Now, either we spend the next 25 years multiplying lawsuits trying to preserve an obsolete advantage, or we start right now reinventing the model—while thanking those 25 years that allowed us to exist.

I understand it upsets a well-established model. But Google isn’t the only culprit. Media must reinvent themselves. Many are already doing so in the United States, with intelligence and inventiveness. They have a promising future.

AI Reveals More Than It Destroys

AI is now designated as the main culprit for the end of traffic and clicks. In reality, it’s mainly the last nail in the coffin—the one that makes the problem truly visible.

To understand the ongoing transition, let’s put ourselves in the shoes of the user searching for something today on Google or OpenAI.

Until 1995, Google was really a search engine operating with keywords: you put 1, 2, 3, 4 words maximum in the search bar. We were even told not to make queries too long to avoid landing on irrelevant sites. You had to target your keyword to get the right pages.

In 2015, Google made one of the deepest updates to its search engine by injecting AI into its algorithm via updates like Hummingbird and Google RankBrain. This AI worked in the background to better detect the intent behind our queries. The funnel was then quite clear: informational, commercial, then navigational.

Since 2022, what changed with AI is that it was injected front-end, directly into the user exchange interface. ChatGPT launched the first arrow. Google followed, moving from a search engine to an answer engine.

What new opportunity?

👀 We’ve moved from a short query composed of a few keywords to a real monologue: the prompt. This (famous) text that can now be extremely detailed that Google or OpenAI will dissect to detect our real search intent.

Consequence: searches have become not only much more contextual, but also much more personal. They are hyper-refinable and detectable from the start. AI optimizes these first two parts of the funnel.

AI Doesn’t Kill the Click, It Qualifies It

The truth is, AI doesn’t kill the click: it exposes it. Yes, 60% drop on desktop, 50% on mobile. It’s real. But change perspective.

The good news:

• You can prove to your boss or client it’s not your fault—everyone is experiencing this drop

• You can adjust your traffic objectives more realistically

• You can justify budgets to explore alternatives

• And especially: a click from AI is about 4.5 times more likely to convert than a traditional click

Why?

Because the user who clicks after reading an AI Overview is already ultra-qualified. They’ve already had their dose of information, compared options, done their checks. When they arrive at your site, they’re “hot,” ready to buy.

American studies confirm it: the website isn’t dead. Blogs, yes, because the info they provided is now furnished by AI. But strategic pages—homepage, product, service, contact—are bouncing back.

👀 This doesn’t mean the website becomes the sole point of contact, but it remains fundamental. You’ll need to be present on many other channels, but these essential pages regain their importance.

Paradigm Shift: From Traffic to Influence

You need to change your perspective.

AI doesn’t kill your traffic, it qualifies your customers for you. Not only can AI bring you traffic (so zero-click search isn’t really so “zero”), but especially it also saves you traffic by filtering out what you don’t need.

👀 I often hear: “how should I recover my traffic?” But why do you want to recover your traffic if they’re visitors who behave like people in a store who look without buying? You don’t want them to look, you want them to buy. AI acts as a filter.

It’s scary, it’s a big change, but it completely modifies the objectives assigned to content. Before, we were obsessed with clicks. Now, we talk about impressions, position in citations, share of voice, Attributed Influence Value (AIV).

A Concrete Example

Before, the query was simple: “Pilates classes in Miami.” Google offered sponsored links and natural results.

When Google started evolving, it already offered some answers directly on its site.

Today, with AI mode, I can tell my life story: “I’m coming back from surgery, I need Hot Pilates of a certain style. I’d like to find in Miami Beach or Surfside, women-only classes, at least 45 minutes.” AI offers an incredibly personalized answer, even suggesting specific studios with recommended instructresses like... Julia!!

From the user perspective, it’s unbeatable.

This is the new Internet. I know you don’t have it in France yet, but it’s coming—Sundar Pichai repeats it in all his interviews. This will be the new Google.

Today, we’re in an intermediate period. Google will probably orchestrate a mix of the three modes: traditional Search, AI Overviews, and full AI mode.

New Indicators

Forget the old KPIs. Make way for impressions, position in citations, share of voice, AIV. Each citation is an influence signal.

👀 Conclusion: AI Overviews inflate impressions and suffocate clicks. But the clicks that survive have real purchase intent. Now, if your conversion rate explodes, traffic volume becomes secondary.

🟣 The New Content Deal

Let’s now move to the more micro-economic aspect of this transformation. Two fundamental pillars structure this new content reality.

👉 First pillar: understanding how AI modes work: Google has shown some transparency on the subject. The day after their annual conference last May, they published a technical document explaining the operation of “query fan out”—the mechanism at the heart of AI modes.

🤫 This transparency is still somewhat hypocritical: while Google explains the theoretical operation, we still can’t track performance in AI modes via Google Analytics 4 which provides no data (suffice to say SEO technicians are seeing red).

Nevertheless, we still have quite a few elements to understand the new mechanisms.

👉 Second pillar: analyzing user behavior: Beyond the algorithm itself, it’s crucial to understand how users interact with these new AI modes. A particularly relevant recent study enlightens us on these new behaviors. This understanding of the relationship between users and algorithm forms the basis for developing new content strategies adapted to this era.

1️⃣ Understanding How AI Mode Works: Query Fan Out

The central mechanism behind visibility in Google’s AI Mode is the query fan-out technique.

👀 Query fan-out is an information retrieval method used by the AI system (like Gemini 2.0 or 2.5) to deeply understand a user’s complex question.

Decomposition: When a user submits a query, AI doesn’t search for that exact string. Instead, it uses a large language model (LLM) to break down the query into multiple more specific sub-queries (called synthetic queries). These synthetic queries reflect various reformulations of the original intent, such as implicit, comparative, related, or historically co-searched terms.

Parallel retrieval: The system executes these synthetic queries simultaneously. It retrieves documents from the search index not only for the original query, but for the entire fan-out of sub-queries, drawing from various sources like the live web, Knowledge Graph, and specialized data.

Corpus creation: This parallel retrieval creates a temporary “personalized corpus” of highly relevant documents, specific to this user and this query.

Synthesis: AI then processes information extracted from this personalized corpus, combining elements from various sources to produce a single complete, synthesized natural language response.

This process takes only a few minutes.

When Elizabeth Reid was presenting AI Overview two years ago, she summarized the promise thus: “Let Google do the Search for you.”

Today, this vision has become reality.

AI does the search work for you: you no longer need to click on different links, reformulate your queries, or navigate between multiple sources.

AI analyzes your initial question and anticipates all the related questions you wouldn’t have thought of, simply by decoding the words of your request.

It builds a complete answer by exploring these multiple dimensions.

👀 This is where a complete paradigm shift for SEO resides. This transformation fundamentally changes the challenge of online visibility. We’re moving from optimization centered on isolated keywords to matrix logic, where you must think in terms of networks of interconnected questions and themes.

👉 We’re moving from optimization centered on isolated keywords to matrix logic, where you must think in terms of networks of interconnected questions and themes.

2️⃣ Understanding User Behavior Toward AI Overviews

This second dimension is based on a major study conducted by Kevin Indig in collaboration with Similarweb. This data, although likely to evolve in the coming months, reveals truly interesting behavioral trends to guide our content strategies today.

Revealing Engagement Rates

70% of users don’t go beyond the AI Overview. So if your content doesn’t appear in the top 30% of this synthesis, your visibility chances become almost zero.

Nearly 90% of users skim the overview without thorough reading, in only about thirty seconds. However, three profiles are distinguished by more attentive reading: 25-35 year-olds, mobile users, and people performing “YMYL” (Your Money, Your Life) queries—health, finance, and high-stakes economic decisions.

Brand as Decisive Criterion

The study reveals a fundamental change in the decision-making process: 58% of users favor recognized brands before even evaluating the response’s relevance. They don’t verify information, they directly trust the brand.

This dynamic radically simplifies decision-making to only 2 steps:

• Do I trust this brand?

• Does it answer my question?

👀 To-Do: Building a brand that inspires trust and reliability should be your top priority.

The Rise of Social and Community Content

Nearly 18% of clicks from AI Overviews go to forums and social platforms, Reddit leading the way. This phenomenon reflects a clear and not new preference: we give more credit to peer-to-peer exchanges than to official information.

YouTube also establishes itself as the big winner of this transition to AI, with strong promotion in results.

About 40% of users open a second link to verify information.

To-Do: AI Overview + Reddit = the winning equation to multiply your conversions by X.

Behavior Segmented by Generation and Device

Data reveals notable differences by age and tool used.

Mobile regains central importance: 54% of users scroll on mobile versus only 29% on computer.

A generational divide also appears clearly: those over 45 continue to check traditional “10 blue links,” attached to 25 years of search habits. The younger ones, on the other hand, stick to the AI Overview without looking further.

High-stakes queries: For high-stakes queries (called “Your Money Your Life”), 38% of users open a second link to verify information, showing more cautious and affinity-based behavior toward these new interfaces.

🟣 Your “Content” To-Do List (Optimization)

👉 Your absolute priority (1): appear in AI results. Becoming visible in AI-generated results must now become your primary objective. This doesn’t mean your website becomes the sole channel—as previously mentioned—but it remains an essential pillar of your digital presence. Your key pages—products, services, contact, about—must be impeccably optimized.

👀 Example: I can confirm all this through our own tests on my community site: AI captures and uses even pages deemed secondary, systematically mobilizing them to contextualize and enrich its responses. The results of this optimization appear quickly.

👉 Your absolute priority (2): be present in the sources that feed AI. This new deal implies strategic presence in all sources consulted by these algorithms. Two fundamental levers emerge: Build and develop your brand authority. Which necessarily involves building your topical authority.

👀 Recognized thematic expertise becomes the foundation on which all visibility rests in this AI-driven ecosystem. How to develop solid topical expertise? Through the quality and depth of your digital content; through the semantic richness of your publications; through generating qualified traffic; and through optimal technical structuring of your site. I develop this below👇👇.

1️⃣ Choose Your Target Entity

Start by precisely determining how you want to be recognized by your customers: define your positioning. Once done, develop your expertise through thematic clusters.

The Old Approach: Keyword-Centered SEO

For years, natural referencing rested on a simple principle: identify keywords, assign them dedicated pages, then publish content optimized for these precise terms. This method was based on literal correspondence between the user’s query and words present on the page.

In 2025, this approach reveals three major limitations:

• Superficial content: narrow articles, vulnerable to competition

• Cannibalization: multiple pages from the same site compete for the same queries

• Blind spots: absence of coverage on sub-themes, persona-specific questions, and real audience issues

The traditional approach was characterized by weak internal cohesion between pages, decisions guided by search volumes and difficulty scores from tools, and monitoring based on page-by-page traffic projections. This strategy requires much effort for limited durability: it generates temporary ranking but doesn’t allow scaling.

The Crucial Nuance: Topic Clusters vs Thematic Approach

While the concept of clusters and topic clusters isn’t new—dating back to RankBrain and Hummingbird era, popularized by HubSpot in the US and Laurent Bourrelly in France with his semantic cocoon—its application in the AI era requires quite a significant overhaul.

Many teams think they’re already applying a thematic strategy by creating topic clusters. Yet most maintain a keyword-centered mentality, which changes everything.

Traditional topic clusters (keyword logic):

• Pillar page covers main high-volume keywords

• Satellite pages treat long-tail variations of these same terms

• Optimization remains focused on exact matching

• Organization improves, but without true thematic optimization

Modern topic clusters (thematic logic):

• Pillar page offers a complete guide to the topic as a whole

• Satellite pages explore in depth key concepts, different personas, related issues, and unique angles

• Coverage includes even zero-volume queries

• The objective is real authority, not just temporary traffic

The pillar-cluster framework can help structure your content, but only if implemented with a thematic-first approach.

The New Approach: Think Thematic, Not Keywords

The key lies in a holistic approach centered on themes rather than isolated keywords. So... your referencing strategy no longer starts with keywords. It starts with your personas, their problems, and topics that truly interest them.

Thematic SEO consists of:

Cover your main topics across personas: different user types, industries, and usage contexts. To treat a topic exhaustively, keywords alone no longer suffice. You must conduct in-depth analysis of your personas and audience: precisely understand their behaviors, their readings, the channels where they consume information, and how they approach each topic.

Invest in “zero volume”: answer your audience’s real questions, even if tools indicate zero monthly searches. This is undoubtedly the most difficult aspect to get clients, bosses, etc. to accept. I know, and I sincerely sympathize. But... it’s an essential point of evolution: Query Fanout exploits implicit queries, those no one would have spontaneously thought of. To truly master a topic in all its dimensions, you’ll often be led to cover unsearched queries: either because no one has thought of them yet, or because they don’t correspond to any natural formulation in your theme.

Offer diverse perspectives: different angles, positions, and unique viewpoints on your topic. To achieve this, collaboration with influencers is fundamental.

Create information gain: provide new data and unique insights found nowhere else

Why This Approach Works in 2025

Because it accounts for Google’s deep evolutions over 1 year:

Entities trump keywords: Google no longer looks for literal matches. It understands concepts and relationships between entities via its Knowledge Graph.

Query Fan Out: Google now transforms a single query into dozens of variations and sub-queries thanks to its language models. A well-optimized thematic page can rank for over 100 long-tail queries without ever mentioning these exact terms.

Content journeys: you no longer answer an isolated query, but build a content system that accompanies different users at different stages of their journey.

Sustainability: a site that deeply treats a topic from all angles naturally attracts quality backlinks and better resists algorithm changes.

Conclusion

It’s a fundamental change in perspective, not just better organization of your keywords.

Choose your target entity and precisely define how you want to be recognized. Build your topical authority by creating interconnected content clusters. In the AI Search era, the only way to ensure lasting visibility is to adopt a change in approach:

➡️ You move from “Which query to target?”

➡️ to “which topic to master?”

👀 This latter approach will ensure you semantic referencing AND entity-based optimization which are today the two new fundamentals of visibility in AI. This implies moving away from strategies exclusively centered on search volume and keywords, to explain the value of zero-volume searches and ultra-targeted content oriented toward personas and pain points.

2️⃣ Be Present in AI Training Data

Several studies—notably those from Profound, Ahrefs, and SEMrush (the most relevant in my eyes)—have analyzed this question and offer us interesting statistics, even if sometimes contradictory.

Create Your Own Community

When you optimize for AI algorithms, you naturally go beyond Google’s perimeter alone to reach multiple platforms. Studies show ChatGPT draws more from Reddit, Google from Quora (and vice versa, with cross-influences), others from LinkedIn...

👀 But beware of misinterpretation!! This doesn’t mean you should be “everywhere”—or be on Reddit just because everyone is on Reddit. A bit on Reddit, a bit on Quora, a bit on LinkedIn.

The real message is elsewhere: AI considers as a reliable source conversations between real people. Which by nature take place on forums like Reddit, Quora, or elsewhere. Being on Reddit is one tactic among others. The strategic approach, however: is finding a space where you’ll host conversations—therefore creating and developing your own community. Build a platform—whatever it is—where people can authentically exchange about your topic. It’s this community dynamic that will fuel your strategy and naturally generate discussions on other platforms.

Wikipedia and Other Authority Sources

Wikipedia, TrustPilot, and other reference sources are indeed among sources consulted by AIs. Should you therefore throw yourself body and soul into creating a Wikipedia page? No. It’s data to keep in mind, nothing more. If it’s relevant and possible for you to be there, do it. If not possible, no big deal. Being cited by a source that itself references Wikipedia produces the same effect.

The important thing isn’t necessarily having your own Wikipedia page—besides, you may not have the required legitimacy—but being present in the ecosystem of reliable sources.

Your Website Remains Central: 50% of Citations

The data is clear: 50% of citations in AI results come directly from websites. Your website therefore remains fun-da-men-tal.

Imperatives for performing content:

• Ultra-niche and exhaustive: cover all possible FAQs, all nuances, micro-themes, hidden intents

• Zero search volumes included: don’t be afraid to address topics your client hadn’t even thought of

• Extreme depth: treat each micro-topic lengthwise, widthwise, and crosswise

The guiding principle: in building your brand, avoid excessive diversification. Favor depth over breadth.

3️⃣ Optimize Your Content for AI

I synthesize here the essential principles, largely documented elsewhere, but I list them here anyway to help better understand the overall logic.

Accessibility and indexing: Your site must be perfectly readable and indexable by AI. The temptation to block robot access to protect your data proves counterproductive if you want to appear in AI Overviews. That’s why the Cloudflare solution—which proposes charging for AI bot access—isn’t as explosive as it seemed when it came out: most actors prefer to authorize indexing to maximize their chances of appearing in Google results.

Strengthen thematic expertise: As the query fan-out principle demonstrates, the more you cover a topic in depth—anticipating the various search intents and related questions of your users—the more you increase your chances of appearing in AI Overviews and other AI modes.

Optimize by relevant segments: Crucial point: AI no longer just seeks to evaluate a page’s overall quality, but identifies and extracts specific segments that best answer each intent detected in the query. The relevance of these micro-contents—targeted paragraphs—becomes determining. Each synthetic query generated by query fan-out must find its optimized answer in your content.

Facilitate AI’s work: Since AI becomes your first “salesperson,” make its job easier. Structure clear syntheses indicating what you want it to remember. Strategically integrate your identity keywords and brand name in this content. For local businesses, systematically mention your location. These elements must irrigate both your informational content and your institutional pages.

Multiply authority signals: Everything related to “citation dignity”—your legitimacy and credibility—must show through in your content: media mentions, expert and influencer contributions, respect for E-E-A-T criteria (Experience, Expertise, Authoritativeness, Trustworthiness). These authority signals considerably strengthen your positioning.

Adopt multimodality: Text no longer suffices. Visuals—infographics, diagrams, illustrations—must be perfectly integrated and readable by AI. Many tools facilitate this approach, which becomes absolutely essential in the current ecosystem.

🟣 New Content Formats and Creation Tools

Contrary to what we hear, blogs, far from being dead, simply deserve to be rethought. I talked about it in detail here.

All content must be articulated around three central pillars: your industry, your products, and your customers. These three axes form the playing field for your AI Search content strategy.

The good news: a multitude of tools is now available to create original content and, in particular, the holy grail of foundational content: original research.

With a well-thought-out strategy upfront, you can create in a few days sector studies that would have previously cost a fortune.

Here are several ideas for you:

Academic Tools and Sources to Exploit

✅ Google is notably preparing the deployment of its Google Trends API, which will provide valuable data specific to your sector. Invest in in-depth research.

✅ Look at university sources. Let me explain: among the sources most cited by AI, two databases deserve particular attention:

ResearchGate: this platform gathers research from around the world and is among the most consulted and cited sources by artificial intelligence.

arXiv: less known but essential, this open archive hosted by Cornell University (an American Ivy League) publishes cutting-edge research. The first document on GEO was published there by students, demonstrating this resource’s value for documenting emerging trends.

The Model Context Protocol (MCP): A Revolution Underway

The Context Protocol transforms Google from a passive search engine into an intelligent executive assistant. Think of it this way: before, Google gave you links to information. Now, it can connect directly to your data and business systems and act on your behalf.

Example of use:

Before MCP: You search “e-commerce conversion rate Q4 2024.” Google lists 10 articles to read. You spend 30 minutes manually compiling your data and a bunch of time cross-referencing.

With MCP: You ask “Compare my Q4 conversion rates to sector benchmarks and identify improvement points.” Google accesses your Google Analytics, your CRM data, and recent market studies. You get an actionable comparative analysis in 2 minutes.

The fantastic impact: less time spent searching and compiling, more capacity to create ultra-personalized content enriched with your own data and ideas, without significant budget.

Available Tools and Brilliant Examples:

Several platforms have developed their own MCP (Model Context Protocol), notably HubSpot, SEMrush, and Ahrefs. Implementation requires some code but remains accessible. And you have at your fingertips a goldmine of data for WOW content!!

“Vibe Coding”: From Content to App

“Vibe coding”—or ambient coding—radically simplifies content and application creation. No longer need to develop line by line: just describe to an AI (like Claude, ChatGPT, or Cursor) what you want for it to generate it.

You say “create me a modern landing page for my SaaS product with a contact form” and 30 seconds later, you have a functional webpage. That’s vibe coding. You give the “vibe,” the intention, and AI builds.

You may not know it, but you’re certainly already using Vibe Coding tools without knowing it. Google has created several in recent months that are absolutely fantastic.

I’ll give you my favorites:

✅ Notebook LLM: the favorite tool for creating practically any type of content. It can replace or complement a blog and now allows generating videos and other formats. I’m an absolute fan. I’ve created tons for all my curation. I’m sharing this one which allows you to know everything about Notebook LLM precisely.

✅ Google Gems and GPTs: community members create remarkable GPTs based on reference works, offering extraordinary possibilities.

✅ Claude’s artifacts: widely adopted for their efficiency.

✅ Google Opal: a gem for juggling from one content format to another (from blog to video via podcast) all starting simply from a sentence that generates a tree structure. Opal being still in testing only in the US, I tested it for my community in a Loom.

🟣 I’m Karine. I help you adapt your content strategy to AI Search and GEO through a masterclass and a community focused on AI marketing. The next session of my GEO masterclass in english will be in May. Want to join?

Join the waitlist

The New Reality of Online Visibility

We are witnessing a fundamental shift in search. Google is no longer just displaying blue links; it generates answers directly via AI. This transformation manifests through two systems: AI Overviews (the AI snapshots that appear at the top of search results) and AI Mode (which operates via a system called “query fan out” where a query breaks down into multiple synthetic sub-queries).

The big difference? In AI Overviews, you optimize to be cited in a summary visible to everyone. In AI Mode, you optimize to be relevant across a much larger and invisible space of synthetic queries. But both share a common reality: traffic will drop massively, and visibility becomes more important than clicks.

Here’s how to adapt your content strategy to survive and thrive in this new era.

1- Immediately Change How You Measure Success

Applies to: AI Overviews AND AI Mode

The first thing to understand is that your traditional KPIs are obsolete. Measuring only traffic and clicks is like measuring CD sales in the streaming era. You’ll panic over declining numbers when what really matters is your visibility.

For AI Overviews, the study is clear:

When an AI overview appears, outbound traffic drops by 66% on desktop and 50% on mobile. And this isn’t temporary, it’s structural. About 21% of users find their complete answer in the overview and don’t click anywhere. They read, are satisfied, close Google. Your site was never visited.

For AI Mode, it’s even more radical since the system doesn’t necessarily even show traditional results. Your content can be used to generate an answer without you receiving a single click.

So here’s what you need to measure now:

• your impressions in AI overviews,

• your position in citations (are you in the first third?),

• and your share of voice (how frequently does your brand appear in AI responses across different queries).

For AI Mode, consider each citation as a signal of your attributed influence, your AIV (Attributed Influence Value). It has become a mechanism for awareness and trust building rather than a direct source of traffic.

Immediately alert your management: traffic will drop by 30 to 50%, and this is not a SEO performance problem, it’s a structural market change. If you depend on display advertising or affiliation, your business model is in critical danger. You need to plan budgets to compensate for this loss and explore alternative revenues like subscriptions, direct services, or proprietary products.

2- Master Thematic Depth and the Latent Query Space

Applies primarily to: AI Mode (but also beneficial for AI Overviews)

AI Mode works differently from traditional search. When a user asks a question, the system doesn’t just try to answer that specific question. It automatically generates dozens, even hundreds of synthetic sub-queries to explore all possible angles of the topic. This is called “query fan out”.

Imagine a user asks “how to choose an electric car”. The system will generate sub-queries like “average electric car range”, “cost of electric charging vs gas”, “best electric cars 2025”, “charging infrastructure by region”, “electric car maintenance”, etc. Your visibility depends on your ability to cover this complete latent query space, not just the original query.

Concretely, you must adopt a “answer a facet” mentality. For any broad topic, think about the various sub-questions a user might explore. Then, build complete thematic authority by exhaustively covering each facet. Don’t create a single 3000-word article on “electric cars”. Create ten detailed articles on each aspect (range, total cost of ownership, infrastructure, insurance, maintenance, resale, comparisons by segment, etc.) and link them semantically to each other.

For AI Overviews, this thematic depth remains relevant but in a different way.

The study shows that users read on average only 30% of the overview (the first third), and 70% never go beyond. So your goal is not to be cited anywhere in the overview, but to be cited in the first three sources. And for that, having recognized thematic authority drastically increases your chances.

Another crucial difference: optimize for conversational queries.

Both systems (AI Overviews and AI Mode) excel at understanding natural language. People no longer type “italian restaurant paris 11th”, they ask “where to find a good italian not too expensive in the 11th with a terrace?”. Your content must be aligned with these naturally expressed intentions, not just stuffed with keywords.

3- Structure Your Content for Machine Readability

Applies to: AI Overviews AND AI Mode (but differently)

AI systems don’t consume your content like a human. They extract passages, structured elements, information fragments.

• For AI Mode, the LLM reasons at the passage level and performs logical steps to construct its response.

• For AI Overviews, Google selects specific excerpts to cite.

In both cases, your content must be built in modular formats that are easy to parse and reuse. Use clear headings, descriptive subheadings, short bullet lists, concise answers to common questions.

But beware: the study on AI Overviews reveals that 86% of users “skim” rather than read in detail. They scan visually to spot answers.

So here’s the optimal structure:

1. each paragraph or section must be semantically complete and capable of answering a specific sub-query by itself. This is critical for AI Mode where the system can extract just a passage without context. But it’s also important for AI Overviews where your cited excerpt must be self-sufficient.

2. Next, be answer-oriented. Content that presents clear, direct answers, particularly at the beginning of a paragraph or section, is much more likely to be used. Create “micro-summaries” in layers: a TL;DR box at the beginning, then the direct answer in 2-3 sentences, then details for those who want to dig deeper. This structure works perfectly for both systems.

→ A specificity of AI Mode: articulate comparisons explicitly. Many AI responses involve decision-making, so content that clearly articulates advantages, disadvantages, and trade-offs (”why X rather than Y”) has a better chance of surviving the LLM’s pairwise ranking and synthesis. For example, instead of writing separately about PayPal and about Wise, write a section “PayPal vs Wise: which to choose according to your situation” with a clear comparative table.

→ For AI Overviews specifically, the study shows that users look for visual signals to quickly find information. So maximize visual hierarchy: bold text for key facts (sparingly), short lists of 3-5 items maximum, generous white space, moderate icons or emojis to mark important points, comparative tables instead of dense paragraphs.

3. And a critical point for both: go multimodal. Invest in varied formats (video, audio, images, infographics). Visibility depends on being the most relevant content in all formats the system can process. The study shows that for “how to” queries, users overwhelmingly prefer YouTube. So if you don’t have a tutorial video, you’re missing a huge opportunity.

4- Prioritize Authority and Trust Signals

Applies to: AI Overviews AND AI Mode

Here’s the most groundbreaking discovery from the AI Overviews study: search behavior has changed radically.

• Before, users wondered “does this result answer my question?” then clicked.

• Now, they apply a two-step filter. First: “do I trust this source?”. If the answer is no, they move on without even checking if it answers their question. If the answer is yes, then only do they check relevance.

Result: 58% of users prioritize a recognized source as a priority, even before relevance. Authority trumps relevance. A study participant expresses it perfectly: “Yelp is a good resource that I use a lot, so I would probably click on Yelp. I trust Yelp.” Another says: “US News and World Report is trustworthy. Edmunds is trustworthy.”

For AI Mode, it’s similar: AI relies heavily on reliable sources to generate its responses. Your goal literally shifts from “getting a higher ranking position” to “being the source that Gemini cites”. And for that, E-E-A-T signals (Experience, Expertise, Authority, Trustworthiness) are absolutely critical.

Concretely, highlight original research, author qualifications, and transparent sources to maximize credibility signals.

For AI Overviews, the study shows that users look for immediate trust evidence. Take the example of NerdWallet which displays seven trust signals just in the header: transparent disclosure about advertising, explanation “how we make money”, author name and photo, reviewer name, editor name, “expert verified” badge, last update date. Result: position number 2 for “best way to transfer money”. So your immediate trust checklist, for each key page:

• identified author with credentials (not just a name),

• publication AND update date,

• expertise badge or mention,

• transparent editorial process,

• visibly cited sources,

• clear disclosures (affiliations, sponsoring),

• accessible contact or support.

For AI Mode specifically, design your content to be “citation-worthy”. Present factual, attributable, and verifiable information. Use named sources, quantitative data, and semantically clear statements that LLMs can extract with high confidence. Avoid ambiguous language or unsourced claims.

And a crucial technical point for both systems: feed the Knowledge Graph !! Use structured data meticulously (like FAQPage, HowTo, Article, Organization, Author schemas with credentials, authentic Review). Specific and clearly named entities must align with Google’s Knowledge Graph to improve your visibility.

5- Build Your Brand Authority Massively

Applies to: AI Overviews AND AI Mode

If you retain one thing from this entire guide, it’s this:

In the AI era, your brand must be known BEFORE users see you in results. This is non-negotiable.

The AI Overviews study is categorical: users now filter in two steps, and trust (therefore brand recognition) precedes relevance.

This means you must invest massively in building your “share of opinion”. Users must know your brand before they encounter you in an AI Overview or an AI Mode response. Invest in public relations, media presence, partnerships with established brands, consistent presence on platforms where your audience is.

A concrete example: Homes.com invested $1 billion in advertising (two Super Bowl 2025 ads, presence at Grammys, Oscars, Olympics). Measurable result: their spontaneous awareness went from 4% to 36% in one year, with a direct correlation between brand traffic and non-brand traffic.

Obviously, you probably don’t have a billion to spend. But the principle remains the same at all scales.

• With a €5,000 monthly budget, you can sponsor niche podcasts, publish guest posts on established media, maintain a consistent presence on LinkedIn with expert content, and work with micro-influencers in your sector.

• With €20,000 per month, add targeted display advertising, industry event sponsoring, active press relations, and regular video content.

• With €100,000 per month, you can consider local TV/radio campaigns, partnerships with established complementary brands, proprietary events, and original research cited by media.

And here’s something crucial → get mentions from absolute authority sources.

There’s a clear hierarchy.

• Government sites (.gov), academic institutions (.edu), and international organizations are at the top.

• Then come major media, recognized professional associations, and think tanks.

• Finally, leading specialized publications and recognized industry experts.

Why is this important?

Because Google cites these sources first in AI Overviews, and AI Mode gives them much more weight in its reasoning.

So … :

• Create citable original studies, offer your expertise for their resources, participate in public consultations, sponsor academic research.

• Develop exclusive data (surveys, studies), position yourself as an expert for journalists, comment on your industry news quickly, propose unique angles on trending topics.

For AI Overviews specifically, there’s a powerful priming effect. If your brand appears in the AI overview, then in Reddit discussions or YouTube where people go to validate, you “win” twice.

The study shows that 18% of outbound clicks after consulting an AI Overview go to community forums (mainly Reddit), with a strong proportion to YouTube videos. And 38% of users open a second link “to verify”. This is your opportunity.

6- Adapt Your Strategy According to Your Audience Demographics

Applies primarily to: AI Overviews (but also relevant for AI Mode)

The AI Overviews study reveals a major discovery: age changes absolutely everything.

• Young people (18-34 years) trust AI overviews more, accept AI as a final answer in 50% of cases on mobile, and seek validation on Reddit and YouTube.

• Older people (45+ years) still prefer traditional “blue links”, look for established authority sites, are more skeptical of AI, and continue to click on classic organic results.

The device factor is also massive. On mobile, users scroll through 54% of the overview on average, versus only 29% on desktop. Mobile also generates more internal clicks in the overview (19% vs 7.4%). Conclusion: mobile has become the priority format, and you must optimize it first.

So segment your strategy.

• If your audience is mostly young (18-34 years), put maximum priority on presence in AI Overviews, invest massively in Reddit and YouTube, create AI-optimized content, focus on social proof. Your budget mix should be approximately 40% AIO and SEO optimization, 25% Reddit/Forums content, 20% video (YouTube, TikTok), 15% brand building on social networks.

• If your audience is mostly older (45+ years), traditional SEO remains king. They prefer blue links, so organic ranking is critical. Authority and credibility are non-negotiable. User experience must be classic: clear and simple navigation, readable text with sufficient font size, no “too modern” design that disorients. AI Overviews remain important but don’t over-index on them. Your budget mix should be approximately 60% traditional SEO and long-form content, 20% brand building (PR, established media), 10% AIO optimization, 10% email marketing (very effective for this demographic).

• If your audience is mixed, segment your pages by intent and probable age. Create differentiated journeys. For example, a “Complete Guide” page in long and detailed format for the older audience, and a “Quick tips video” page in short and visual format for the young audience. Same topic, different formats, different audiences.

There are still exceptions where intent trumps demographics.

• For local searches, all ages use local packs, so Google Business Profile optimization is critical for all.

• For online shopping searches, all ages compare online before buying.

• For YMYL searches (health, finance, legal), all ages seek multiple validation and authority sources are critical for all.

7- Understand the Impact of Perceived Risk

Applies primarily to: AI Overviews

The study shows that for high-stakes searches (health, finance, legal), users show more hesitation and confusion, adopt a “double-checking” behavior, and 38% of sessions include opening a second link “to be sure”. They actively seek authoritative sources. For low-stake searches, it’s the opposite: quick acceptance of the AI overview, little or no verification, quick decision.

Translation: the risk level determines behavior. High risk equals more clicks, therefore more opportunities for you. Low risk means the AI overview suffices, therefore fewer clicks.

• If you’re in a YMYL sector (Your Money Your Life: health, finance, legal), here’s good news: you’ll lose less traffic than other sectors. Users still verify, look for multiple sources, and trust is non-negotiable. Your specific strategy: display massive trust signals (visible professional certifications, institutional affiliations, clear editorial review process, authors with verifiable credentials), get mentions from .gov and .edu sources (these sources are prioritized in AI overviews and users actively seek them), and optimize for “verification” (your content must be findable when someone wants to verify, with an ultra-clear title and meta description about expertise).

• If you’re not in a YMYL sector, the bad news is your traffic losses will be greater. To compensate, focus on speed and convenience (users want a quick answer, optimize loading time, simple and direct interface), display competitive visible pricing (for e-commerce: clear price from the SERP, visible promotions, price structured data), and differentiate through brand (since trust weighs less in your sector, innovate on product and unique user experience).

8- Optimize for “Verification Clicks” and Multi-Platform Validation

Applies primarily to: AI Overviews

Here’s a crucial insight from the study:

Only 19-22% of users finish their search with the AI overview. But 38% open a second link “to verify” or “be sure”. The AI overview doesn’t kill all clicks, it creates a new type of click: the validation click.

This is your new opportunity.

These validation clicks go primarily to Reddit (18% of outbound clicks), YouTube (strong proportion), and authority sites.

Users adopt predictable behavior: they read the AI overview, the overview gives an answer, they want other humans’ opinions, they click on Reddit or YouTube to confirm.

So your strategy must be multi-platform.

1. First, build immediate trust on your web page to capture these verification clicks (follow the checklist mentioned earlier).

2. Next, drastically reduce “Time to Value”, meaning the time between arriving on the page and getting the answer. The answer must be visible without scrolling, loading time must be under 1.5 seconds, no pop-ups that block content, no intrusive ads before content.

For Reddit specifically, establish an authentic organic presence. Identify relevant subreddits for your sector, participate authentically (no spam), answer questions related to your expertise, create value before promoting. Set up active monitoring: use tools to track mentions of your sector, respond quickly to questions, offer help for free. If you’re cited in an AI overview AND mentioned on Reddit, it’s a winning combo: both reinforce each other.

For YouTube, optimize thumbnails (users decide from the automatic preview, clear thumbnail that shows the final result, text readable even in small format). The first 15 seconds are critical: immediately show the action or result, no long introduction, the user must see that you answer their question. For “how to” queries, users often ignore the AI overview because they want a video directly. Visual demonstration beats textual explanation.

And don’t forget Amazon if you’re in e-commerce. The study shows that Amazon is a major competitor to Google’s shopping carousel, and many users say “I would just go to Amazon”. If you sell products, an optimized Amazon presence is mandatory, with numerous reviews and ratings.

9- Adapt Your Approach According to Competing SERP Features

Applies primarily to: AI Overviews

AI overviews aren’t the only elements that attract clicks.

• For shopping searches, 30% of clicks go to local packs, 26.4% to shopping modules, and 13.2% to text ads. Result: 70% of clicks bypass classic organic links.

• For tutorials, videos are hyper-priority with an average visit time of 37 seconds (versus 31 for the AI overview).

→ So for local searches, optimizing your Google Business Profile is imperative. Invest in quality photos, get numerous and recent customer reviews, respond to reviews, publish regularly.

→ For product searches, optimize your Google Shopping feed (high-quality images, competitive prices, product reviews), and test ads strategically (13.2% of clicks go to ads, so for commercial queries advertising remains relevant).

→ For tutorial searches, adopt a video-first strategy: for any “how to” query, create a video, optimize for automatic preview, clear demonstration from the first seconds.

In Conclusion: From Keyword Optimization to Relevance Engineering

The transition from traditional search to the AI era requires a complete paradigm shift. You can no longer simply optimize for keywords and hope for traffic. You must now practice what’s called Relevance Engineering: anticipate the synthetic query space (for AI Mode), optimize for visibility in citations (for AI Overviews), build massive brand authority (for both), and diversify your traffic sources beyond Google.

The phrase to remember:

“If your brand doesn’t appear in the first third of an AI Overview, it is effectively invisible. Search has moved from a click economy to a visibility economy, where authority now surpasses relevance.”

Start this week by:

• changing your measurement KPIs,

• alerting your management to inevitable traffic decline,

• and doing an urgent audit of your positions in AI overviews.

This month:

• rewrite your priority content with concise answers at the top,

• launch your public relations campaign,

• and establish your presence on Reddit and YouTube.

This quarter:

• do a complete E-E-A-T technical overhaul,

• launch your proprietary newsletter,

• and explore new revenue models.

Over 6-12 months:

• become an absolute authority in your field,

• build a coherent brand ecosystem across all channels,

• and work towards Google independence with 50% or more of your traffic coming from non-Google sources.

The future belongs to brands that understand that visibility has become more important than traffic, and that authority now trumps relevance.

About the Author

Karine Abbou is a Content Strategist and AI Search Educator. She is also the founder of the Women Love AI Marketing community, a community dedicated to supporting Marketing Professionals and Executives in their transition to AI Search.

💼 LinkedIn: linkedin.com/in/karine-abbou-phd-62318122
🌐 Community: womenloveaimarketing.com