According to Graphite’s CEO, AI Search is not killing SEO at all. It’s making the playing field bigger
Call it AEO, call it GEO: the dispute over names is, in the end, the least interesting part of a very real transformation.
What matters is the concrete fact: today, ranking on Google is no longer enough. Large language models like ChatGPT and Perplexity are becoming answer engines in every sense, and companies are starting to ask themselves how to become these systems’ preferred answer, not just a link in a SERP.
Around these topics I see a great deal of confusion, fueled by weak data, commercial narratives and widespread biases. The result is an ecosystem where the “end of organic” panic spreads faster than the actual data.
As you know, our SEO Confidential column was born precisely for this reason: to bring some order back among often unjustified alarms and fashionable but hollow hype.
Today’s guest is Ethan Smith, CEO of Graphite, an SEO (and AEO) agency working with some of the fastest-growing companies in the world: MasterClass, Robinhood, Webflow, Notion. More than a market observer, Ethan is someone who works in this field every day, with access to large-scale data and the ability to read real user behavior, beyond the myths (often crafted on purpose out of self-interest) that circulate on social media.
We have heard plenty of dystopian stories and theories in recent months. Graphite chose to measure and refute them. What emerges from their research scales down the dominant narrative: the decline in organic traffic exists, but it remains contained, and the main causes appear to be tied to Google’s algorithm updates rather than to the rise of artificial intelligence.
But in what sense would the collapse of organic traffic be a myth? What does Ethan mean, and what data would prove it? And why, according to Graphite’s CEO, is AEO a more accurate term than GEO?
Read our interview to find out.
The novelty bias strikes again! Here are the Graphite numbers that debunk the alarmism
I read with great interest the Graphite study you published recently, which shows that the state of organic traffic is, after all, not so dramatic. How do you explain such a wide discrepancy between aggregate data and the widespread perception of an organic traffic collapse?
Several studies report dramatic declines based on self-reported surveys. That is not the best methodology, because when you try to quantify past behavior, inaccuracies can be frequent.
There is also an incentive for the new AI players to claim that a major discontinuity is underway in the world of online search. That narrative works better than one where nothing is changing, because of the so-called novelty bias, the tendency to give more weight to whatever looks new. We have been hearing that search is dying for a long time: this is just the latest in a long series of similar predictions. First it was Amazon, then social media, then mobile apps. First comes the novelty bias, then people look for confirmation of their own beliefs (confirmation bias), and then they stop there.
Then there is the matter of the commercial narratives around AEO and GEO. There can be a preference for clean, easy-to-communicate disruption stories, like “AI replaces Google”, over more complex coexistence scenarios. Emphasizing a supposed market disruption is, moreover, a very effective strategy for attracting investors and clients.
While artificial intelligence is indeed driving a major transformation and the growth of new channels, that does not mean traditional search has simultaneously declined. Search is evolving alongside AI, but saying SEO is dead is not just an exaggeration: it is an inaccurate statement. Traditional search is not declining. On the contrary, adding together search engine usage and AI query sessions, overall search usage grew 26% globally and 16% in the United States comparing 2025 to 2024. In other words, the pie is getting bigger.
The report shows that AI Overviews significantly reduce CTR, but only on a portion of SERPs and mostly on informational queries. Still, doesn’t this scenario risk compressing the economic value of SEO, concentrating it on an ever smaller number of keywords and business models?
It would be a risk if search demand as a whole were shrinking, but that is not what the data show. What we are witnessing is a reallocation of clicks, not a collapse in value. Yes, there is greater concentration on higher-value keywords. CTR drops 35% when AI Overviews appear, and they show up on roughly 30% of SERPs, which is why we observe an overall traffic decline of 2.5%. That is not a dramatic change, though.
AI Overviews mostly appear on informational queries that have historically always had low or indirect monetization. These have always been top of funnel queries, useful for generating awareness, but rarely representative of the economic core of most companies. At the same time, transactional and commercial queries remain largely intact. There is no compression of total value, but rather a concentration from lower-value searches toward higher-value ones.
In the white paper you question the reliability of surveys, limited samples and isolated cases. In a phase of great uncertainty like this one, how much does the way data is communicated contribute to generating panic or unrealistic expectations in the SEO industry?
A substantial part of the panic comes precisely from the way weak data is presented as if it were certainty. Many of the most alarmist claims are based on self-reported surveys, in which people are asked to estimate how often they search or use AI tools. It is an inherently unreliable method.
When you compare this approach with large-scale panel data, collected across tens of thousands of sites, the picture looks very different: overall organic traffic is down by about 2.5%. Communicating survey-based estimates as if they were objective, definitive results fuels overblown narratives that do not hold up against real data on user behavior.
The debate around GEO, LLMs and SEO seems to have become polarized: on one side there are those who talk about the end of organic search, on the other those who deny any change, fearing imminent bubbles. How was the Graphite report received? Did you get any criticism or particularly hostile reactions, even unexpected ones?
The response was largely positive, with many people sharing how what we published mirrors what they are also seeing in their own data and their own operations.
One of the issues that emerged most strongly is that zero-click answers may drive an increase in branded searches on Google, an aspect this study does not take into account. Many users, after reading an answer provided by the AI without clicking, end up searching for the product or brand directly on Google. This behavior was not included in the analysis and, for now, it is not yet clear what the best method is to measure it rigorously.
Despite the massive spread of AI-generated content, it would seem that search engines and answer engines still largely favor content written by humans. This finding appears to contradict many aggressive editorial strategies: are we witnessing an AI rush unsupported by real results?
Many companies are publishing millions of AI-generated articles in an attempt to capture traffic from Google Search and from answer engines like ChatGPT. Numerous case studies claim this is an effective strategy.
In our study titled “How Does AI-Generated Content Perform in Search and Answer Engines?”, however, it emerges that, despite the now pervasive presence of AI-produced content, it generally does not perform well in search engines or in answer engines:
- 86% of the articles ranking on Google Search are written by humans, while only 14% are AI-generated;
- 82% of the articles cited by ChatGPT and Perplexity AI are written by humans, versus 18% of artificially generated content;
- when AI-produced articles do appear on Google Search, they still tend to rank lower than content written by human authors.
Topical Authority significantly accelerates organic visibility, while the market keeps pushing toward fast, massive content production, don’t you think? But aren’t we running the risk of optimizing for the wrong kind of speed, sacrificing structural results in the name of short-term shortcuts?
Producing high-quality, genuinely useful content is undoubtedly the right choice. Doing it quickly and at scale can be just as valid. What deserves the most attention is, above all, the choice of which content to produce.
Topical authority can help set priorities, identifying the topics where you have the best chances of being competitive. If that selection work is done effectively, there is nothing wrong with moving fast. Speed and mass content production are not a problem in themselves, as long as they are backed by a truly effective strategy.
In the debate between AEO, GEO and AI SEO there is a strong focus on the name, but also on what that name implies for the future of the channel. How much does the label really matter compared to the risk of misunderstanding the nature of the ongoing change, especially for non-technical companies and decision-makers?
Ultimately, the meaning of a term is whatever a group of people collectively decides it is. That is why it will be the market itself that determines which definition prevails.
AEO and GEO are, in fact, trying to describe the same phenomenon: how to make a brand or a piece of content appear in large language models as the answer to the user’s question. Personally, my preference goes to Answer Engine Optimization over GEO, because the term “generative” can also refer to the generation of images, videos and other content that does not necessarily coincide with a textual answer.
The term “answer”, on the other hand, is narrower and describes more precisely the goal marketing focuses on: promoting the brand and gaining visibility inside AI-generated answers.
In your work you dismantle, point by point, the thesis that search is in decline, and yet part of the industry keeps talking about the “end of organic” as an established fact. At this point the question is inevitable: are we facing a collectively distorted reading of the data, or a narrative deliberately built because it serves those selling new solutions, new services and new fears?
What is happening looks a lot like a classic zero-sum interpretation, a well-documented cognitive bias that leads people to think gains in one area must necessarily correspond to losses in another. The underlying assumption is simple: if AI usage grows, then search must inevitably shrink.
It is a pattern we have seen before. When mobile apps started spreading explosively, many argued the web was destined to die. In 2010 even Wired published the famous “The Web Is Dead” thesis. Mobile apps did indeed grow extraordinarily, but the web did not collapse at all. Today the exact same mechanism seems to be repeating itself.
Monthly AI sessions have now reached 56% of global search volume, and yet search has not declined. Previous projections that only looked at ChatGPT’s web traffic underestimated actual AI usage by a factor of 4 to 5, given that 83% of AI usage happens through mobile apps and not via browser.
So we are facing a combination of factors: on one side, a narrative strategically built by those selling AI-based software and services; on the other, professionals and companies acting in good faith who have recorded a traffic drop and are looking for a plausible explanation for what happened.
Ethan, let me play devil’s advocate for a moment: even if SEO traffic were down “only” 2.5%, aren’t we still witnessing a structural loss of organic value, with fewer and fewer real clicks and more and more answers kept inside Google and the LLMs? Isn’t there a risk of downplaying a problem that, economically, many publishers already perceive as very concrete?
There is indeed a real reduction in click-through rate when AI Overviews appear, especially on informational queries. However, these are queries that, in most cases, were not the main revenue drivers to begin with. Transactional queries, the ones most directly tied to conversions, are far less affected.
The point, then, is not that fewer and fewer people are using Google. Rather, Google is directly answering a growing number of informational questions. From this perspective, the macroeconomic impact looks more contained than many headlines and alarmist narratives suggest.
That said, at the microeconomic level the story changes. Individual sites have recorded much sharper traffic drops, and that significant decline is most likely less related to a reduction in search demand and more attributable to the algorithmic changes introduced by Google.
In your talk for Brighton SEO you openly criticize AEO startups and consultants who talk about the collapse of search to legitimize their own positioning. But isn’t there a risk that the opposite position, namely “nothing serious is happening, everything is stable”, ends up reassuring the market too much and delaying an adaptation that will, in any case, be painful for many business models?
SEO has not collapsed, and the data does not support narratives of a substantial decline. At the same time, it is clear that AI-based search is growing and changing how visibility is earned.
Saying SEO is not declining is not the same as claiming nothing is happening. It means, rather, that the channel remains strong, but the conditions for success are evolving. It is important to separate the substantial stability of search from the growth of AI.
AI growth is adding to search, not replacing it. Artificial intelligence now records 45 billion monthly sessions globally, and ChatGPT alone accounts for 20% of search-related traffic worldwide. Yet search engines have not declined. In other words, the pie is expanding. Overall search usage, adding together search engines and prompts to AI systems, has grown 26% globally since the launch of ChatGPT.
If in two or three years organic traffic were to remain stable, but LLM intermediation had drastically reduced publishers’ branding, margins and ability to monetize, wouldn’t we risk being right on the numbers but wrong on the direction of the market? In short, aren’t we perhaps confusing the survival of the channel with its actual economic sustainability?
Since answers are often zero-click, there is a concrete risk that LLMs come to control an ever larger share of the user journey, leaving publishers with less and less control. It is a phenomenon that, in part, we have already been observing in search for years, with Google tending to retain more and more searches and keep users inside its own ecosystem.
The net effect on search currently stands at -2.5%, a relatively contained figure. However, the risk that both search engines and LLMs progressively keep users inside their own products, sending less and less traffic to publishers, remains concrete and should not be underestimated.
AEO or GEO? The only certainty is that you need to become the AIs’ answer!
The SEO industry has historically been fertile ground for doom prophecies. First mobile apps were going to kill the web, then social media was going to empty out Google.
Every time, the novelty bias comes first, the fascination with the new thing that sweeps everything away, and then confirmation bias turns every weak signal into confirmation of an imminent collapse.
Graphite’s data invites us to take a step back: overall search usage has grown 26% globally, organic traffic is down 2.5%, transactional queries remain substantially intact. It is not a collapse: it is a reallocation.
The most alarmist narratives, often serving those who sell new solutions, do not hold up against real user behavior.
My reading is that AI is adding to search; not replacing it. In short, the pie really is getting bigger.
The real risk, however, is a different one, and it is worth naming it precisely. When a user gets an answer directly from Google or from an LLM without clicking, traffic to publishers decreases, even if the overall volume of searches grows.
This means publishers may keep being cited, referenced, even recommended, and yet progressively lose the direct relationship with their own audience. And without that direct relationship, you lose the ability to build a brand, retain readers, and monetize sustainably.
Traffic is a number. The relationship with your audience is a strategic asset. Confusing the two is the mistake the industry cannot afford to make in the coming years.
My heartfelt thanks to Ethan for this stimulating conversation, for bringing us solid data and critical thinking, and for his ability to “swim against the tide” in an industry that often prefers alarm over precision.
If you work in digital, marketing or publishing, these are exactly the kind of insights to keep in mind in the coming months. Because the noise out there will grow, catastrophic predictions will multiply, and staying anchored to real data will make more and more of a difference.
See you in the next episode of SEO Confidential.
#avantitutta
