If you spend money on PR, there is an awkward possibility worth checking: the coverage you are working hardest to get may not be the coverage that decides whether an AI assistant recommends you.
Muck Rack compared the journalists PR teams pitch most often against the journalists AI engines actually draw on for a given brand. The overlap averaged 2%.
That is not a small inefficiency. It suggests most PR effort is aimed somewhere other than where AI systems are looking — which matters more every year, as more of your customers start their search by asking an assistant rather than typing into Google.
Coverage You Didn’t Pay For Is Doing Almost All the Work
Start with the headline finding, because it is stark.
Muck Rack’s continuing study of what AI systems cite — more than 25 million links across ChatGPT, Claude and Gemini, covering seventeen industries — found that earned media accounts for 84% of all AI citations, while paid and advertorial content accounts for 0.3%. “Earned media” means coverage you did not pay for and cannot control: a journalist’s article, a trade publication’s round-up, an analyst’s assessment. The figure has held between 82% and 89% across three editions of the study since July 2025.
Nought point three per cent for everything you can simply buy.
Academic work points the same way. Researchers at the University of Toronto took 300 consumer-electronics queries and classified where each system’s sources came from. Claude drew 65% of its sources from independent media and review outlets, GPT-4o 57%, Perplexity 50%, Gemini 46% — against Google’s 41%. For the queries that matter commercially — best-of lists, comparisons, “which should I choose” — the AI engines converge on earned sources at between 59% and 86%.
So when someone asks an AI which supplier to use, the answer is assembled largely from what other people have written about you. Not from your website.
Why Editors Count for More Than Your Own Pages
The mechanism is not mysterious. An AI system has no way to verify what you say about yourself. It can check whether anyone else says the same thing.
When a trade publication names you in a piece, the decision to include you was made by someone with no stake in your marketing. That is evidence of a kind your own website cannot produce, however well written it is.
Ahrefs measured this across 75,000 brands, looking at which signals track with a brand’s visibility in Google’s AI Overviews. Mentions of a brand across the web scored 0.664. Backlinks — the currency of search optimisation for two decades — scored 0.218. Ahrefs is careful to say this shows association rather than proven cause, and so are we. But the direction is consistent with everything else here: being talked about tracks visibility more closely than being linked to.
Which Outlets, Not How Much Coverage
This is where that 2% figure becomes useful rather than merely alarming.
It does not mean PR does not work. It means most PR programmes are aimed at publications chosen for reach, prestige or relationships — sensible criteria that happen not to be the ones AI systems respond to.
AI citation is concentrated. In the categories the Toronto researchers examined, the sources AI systems leaned on were overwhelmingly specialist review and comparison publications rather than the national titles a PR brief instinctively reaches for.
Knowing which publications carry weight in your category turns a vague “more coverage” brief into a targeted list. That is a more valuable piece of information than an extra month of outreach.
It Has to Keep Going
Coverage decays.
Muck Rack found half of all AI citations come from articles published within the last eleven months, and about 4% from the previous week alone. The Toronto researchers measured median source ages by engine: Claude 62 days, GPT-4o 80, Perplexity 90, against Google’s 130 — AI engines drawing on material roughly 31–70% newer than Google does.
That does not make older coverage worthless; the remaining half of citations spans roughly three further years, and long-lived reference pages do a lot of quiet work. But a campaign you ran eighteen months ago is not carrying you now. This is a drumbeat, not a launch.
Being Cited Is Not the Same as Being Named
There is a complication worth understanding before you judge whether any of this is working.
A large share of AI citations are what the industry calls “ghost citations” — your page is linked as a source, but your brand name never appears in the answer the customer reads. A study of 3,981 brand appearances across four AI engines found that 74.9% drew a citation but only 38.3% got the brand named in the answer text. In 61.7% of appearances the link was there and the name was not — which is 82% of all the citations earned.
The practical consequence is uncomfortable. Your content can be doing the work of informing the answer while a competitor gets the recommendation.
The fix is partly structural. Coverage you secure should carry your name in the sentence a machine would extract, not buried in surrounding context — which in practice means comparative and evaluative pieces, where naming is intrinsic, rather than purely informational ones.
What This Means for You
If AI assistants are not recommending your business, the most likely fix is not on your website.
Three things follow. First, the question worth asking your PR agency is not “how much coverage did we get?” but “which publications in our category do AI systems actually draw on, and are we in them?” Those are different questions with different answers. Second, budget for continuity rather than campaigns — a steady presence beats an annual spike, because the systems weight recent material heavily. Third, check whether the coverage you already have names you in a way a machine can extract, or merely links to you.
None of this requires a large budget. It requires aiming at the right targets, which is mostly a matter of knowing what they are.
Where you stand can be measured rather than guessed at: whether AI tools name your business when someone asks for what you do, how that compares with your competitors, and what in your existing online reputation is producing that answer.
If the pages on your website haven’t been touched in a couple of years, AI tools are quietly less likely to use them when someone asks for a recommendation in your line of work.
That sounds like an argument for publishing more. It isn’t. The research points somewhere more useful and a good deal cheaper: the pages you already have, kept up to date, do more for you than a stream of new ones. Most businesses have this backwards, which is why it’s worth setting out what the evidence actually shows.
“Publish fresh content” has been the standing advice for two years now. What nobody attached to it was a number. Fresh compared to what — last month, last year, three years ago? Without that, it’s a vague instruction rather than something you can act on. There are now numbers, from three large independent studies, and they tell a more interesting story than the advice does.
AI tools draw on noticeably newer material than Google does
Researchers at the University of Toronto measured how old the pages were that each system actually drew on when answering buying-type questions. They ran the same set of questions through Google and through three AI assistants, then checked the publication date of every page each one used.
The typical age of the material they leaned on:
Where the answer came from
Consumer electronics
Cars
Claude
62 days
148 days
ChatGPT
80 days
162 days
Perplexity
90 days
217 days
Google
130 days
493 days
Read across the bottom row first. In a slower-moving market like cars, the typical page Google surfaced was nearly sixteen months old. The AI assistants were working from material five months old. Across both markets, AI answers drew on content between roughly a third and 70% newer than Google’s.
The gap matters because of what it implies about the competition. Older pages aren’t banned — but they’re up against a pool of material that skews much younger than the one you’re used to competing in on Google. And the slower your sector moves, the wider that gap gets. If your competitors are sitting on pages they last touched in 2023, that’s an opening rather than a threat.
Most content AI draws on is still years old
Here’s the finding that stops this becoming a panic about publishing schedules.
Ahrefs looked at nearly 17 million links cited across seven different AI and search platforms. On average, the pages AI tools drew on were 2.9 years old. Google’s ordinary results averaged 3.9 years. So yes — AI leans newer, by about a year. But the typical page it uses is still the best part of three years past its publication date.
Both findings are true, and the difference between them is worth understanding rather than averaging away. The Toronto study measured the midpoint age in two fast-moving consumer markets where publishers refresh constantly. The Ahrefs figure is an average across every topic, and a handful of very old, very authoritative pages pull an average up in a way they don’t pull a midpoint up.
What survives both readings is the same: recency helps, and it helps more than it used to — but a three-year-old page that genuinely answers the question is still being used every day. Age alone doesn’t disqualify you.
Updating an old page counts as fresh
This is the finding that changes what you should actually do on Monday morning.
Seer Interactive tracked 7,683 pages that AI tools drew on across roughly 47,000 mentions between March and June 2026. Three in four of those pages had been updated within the previous year. Judged by when they were last updated, nearly three-quarters looked current.
Then they looked at the same pages by original publication date. On that measure, fewer than half were recent. More than a quarter of the pages that qualified as “fresh” had first been published more than two years earlier.
In other words: most of the fresh content AI tools are using is not new content. It’s old content that somebody maintained.
That is a much easier instruction to follow than “publish more”. You don’t need a content team. You need to know which of your existing pages matter — your main service pages, your pricing or process explanations, the questions customers actually ask you — and to keep them genuinely current.
The word “genuinely” is doing real work there. Changing a date stamp without changing anything else is not an update, and the author of the Ahrefs study warns explicitly against that. What counts is substance: a price that’s right, a service you no longer offer removed, a new question answered, a case or example from this year rather than three years ago.
How fresh is fresh enough?
There’s no single threshold, and anyone who gives you one is guessing. But the evidence supports a reasonable rule of thumb.
Within the last year is the broad working window. Of the pages AI tools drew on in the Seer study, Google’s Gemini took 78% from the previous twelve months, ChatGPT 73%, and Perplexity 65%. Separately, Muck Rack’s analysis of more than 25 million links across ChatGPT, Claude and Gemini found that more than half of the news coverage cited was published within the past year, with the volume dropping off sharply after the first six months.
So: for the handful of pages that genuinely matter to your business, once a year is the minimum, and twice is better. For a fast-moving sector, or for a page carrying prices, availability or anything else that dates quickly, work to a shorter cycle. That’s it. It doesn’t need a dashboard.
What this means for you
If you have been feeling guilty about not publishing enough, you can stop. The more useful question is whether the pages you already have are still accurate, and when you last checked.
Pick the five or six pages that would matter most if a prospective customer read them tomorrow. Look at when each was last meaningfully changed — not restyled, changed. If any of them describes a price you’ve since altered, a service you’ve stopped offering, or a world that has moved on, that is a page currently working against you in two places at once: with the customers who read it, and with the AI tools deciding whether to put you in front of the next one.
There’s a wider point behind this. Whether AI tools recommend your business isn’t a mystery and it isn’t luck. It can be measured — how often you’re named when someone asks for a business like yours, how that compares with your competitors, and what in your wider online reputation is driving the answer either way. Content age is one input among several, and on its own it will never be the whole story.
Finding out where you currently stand is what we do.
Sources: Mahe Chen, Xiaoxuan Wang, Kaiwen Chen and Nick Koudas (University of Toronto), “Navigating the Shift: A Comparative Analysis of Web Search and Generative AI Response Generation”, EDBT/ICDT 2026 Workshops; Ryan Law, “Do AI Assistants Prefer to Cite Fresh Content?”, Ahrefs, July 2025; Seer Interactive, “Content Recency’s Impact on AI Visibility in 2026”, July 2026; Muck Rack Generative Pulse, “What Is AI Reading?”, May 2026.
One of the most common reasons a business looks invisible to AI is not the business at all. It is the question. Ask an assistant about your county and you can look completely absent. Ask it about your locality — the town, the district of a city, or the cluster of villages a customer would actually name — and there you are: same business, same day, same website. The only thing that changed was the size of the map you drew around the question.
By locality we mean the local area a customer would actually say out loud — usually a town, but just as often a named district of a larger city or conurbation, or a group of small, close-together villages that people treat as one place. It is the level below the county or region, and above the single street.
If you are going to check whether AI recommends you, this is the first thing to get right. Get it wrong and you will either panic over a problem you don’t have, or miss a real one.
What we tested
In a dry run of our AI Visibility Check, we took a set of local businesses in one large county and asked the same kind of question three different ways: [1][2]
County level — “best [businesses] in [the whole county]”
Locality level — “best [businesses] in [the firm’s own town or district]”
Auto — letting the tool decide the area for itself from the business’s website
A large county is a good test because it is big — big enough that a customer at one end is not choosing between their local firm and one an hour away at the other end. [3] So “best in the county” and “best in my town” are not the same question at all.
What happened
At county level, not a single one of the fourteen businesses was named. [4] Zero. Asked to name the best across the whole county, the assistant simply didn’t reach any of them.
At locality level, more than half were named — 21 of 38 businesses in the larger sample. [5] The businesses hadn’t changed. The question had.
And when we let the tool pick the area itself, it leaned towards the broad end: for fourteen businesses it chose the whole county six times, the actual locality only five times, and twice — for firms on the county’s edge — it decided they were in the neighbouring city instead. [6] Left to its own devices, it named under a quarter of them. [7]
Why a county question makes you invisible
A county is not a market. “Best in the county” asks the assistant to pick a tiny handful of standouts for an enormous area, so almost every ordinary, perfectly good business gets left off — not because it is weak, but because the question is far too wide for it to be a sensible answer. Narrow the question to the locality a customer would actually search, and the same business becomes a reasonable thing to name.
This is why a scary-looking result can be meaningless. “We’re invisible across the whole county” is nearly true for almost everyone, and tells you very little. “We’re invisible in our own locality, where a customer would look” is the result that actually matters.
The trap hiding in quick checks
Here is the practical catch. Automated “AI visibility” checkers often have to guess your area, and many default to the broadest one unless your website tells them otherwise. If your site doesn’t clearly state the area you serve, a checker can end up testing you at county level — and hand you an “invisible” verdict that a locality-level question would have overturned. In our run, the tool even placed two genuine businesses in the wrong city because their location signals were ambiguous. [6]
So an automated check is only as good as the area it picked. If nobody looked at that, the number underneath it may be answering the wrong question.
What to do about it
Check yourself at the level a customer searches: your locality, not your county or region. If a tool or a report shows you a county-wide “invisible,” re-ask about your own town or district before you conclude anything.
State your service area in plain words on your website — the locality you serve, spelled out: the town, the part of the city, or the villages you cover. This is the single clearest signal that stops an assistant, or a checker, defaulting you to the wrong place (or the wrong city entirely).
Ignore county-level scores as a measure of your visibility. They are a fair test of who the region’s few headline names are — not of whether you show up where your customers are looking.
This is exactly why our own checks fix the question to a business’s actual locality. The county view isn’t wrong; it is simply answering a different question — and it is not the one your customers ask.
Most AI visibility advice is written as though “AI search” is a category with a shared logic. It isn’t. The overlap data draws a sharp hierarchy — and where you sit in that hierarchy is determined almost entirely by which retrieval infrastructure each platform is built on, not by how good your content is.
The Hierarchy Is Not a Coincidence
A Semrush study of 5,000 queries, surfaced in Local Falcon’s April 2026 analysis, maps the overlap between each AI platform’s cited domains and traditional organic search results: Perplexity sits at approximately 91% domain overlap with organic, Google AI Overviews at 86%, Google AI Mode at 51%, and ChatGPT at the bottom — no specific percentage published, but consistently the weakest (source: wiki/localfalcon-ai-search-overlap-2026.md).
The order is not random. Perplexity is built on live web retrieval, so its results track closely with whoever is already ranking. Google AI Overviews and AI Mode draw from Google’s own index — the same one your organic SEO has been feeding for years, though AI Mode shows meaningfully less alignment than AI Overviews, which is its own strategic wrinkle. ChatGPT operates through a fundamentally different retrieval pathway with minimal Google overlap, confirmed by peer-reviewed academic data: a University of Toronto study measuring Jaccard similarity between ChatGPT and Google’s top-10 results found a score of just 4.0% (source: wiki/arxiv-web-search-vs-genai-2601-16858.md).
That 4% is not a gap in content quality. It is a gap in infrastructure.
Where the Brief Goes Wrong
“Optimise for AI” is not a brief. It is a category error.
If you are already ranking well organically, you are almost certainly appearing in Google AI Overviews. The 86% domain overlap means strong traditional SEO transfers almost directly to AIO visibility — these are not separate problems requiring separate programmes. The gap between Google AI Overviews and Google AI Mode (86% vs 51%) is more interesting: AI Mode draws on retrieval differently, and the 13.7% overlap between AIO and AI Mode citations means that even within Google’s own product family, what gets cited diverges sharply (source: wiki/ai-ranking-factors-brandfeatured.md).
ChatGPT is an entirely separate problem. And for most brands, it is the problem that isn’t being worked on.
The citation source pools confirm why. ChatGPT’s top citation source is Wikipedia at 40% of all citations. Perplexity’s top source is Reddit at 50%. Microsoft Copilot leads with Forbes at 32%. Google AI Overviews is the most distributed — NerdWallet tops the list at just 17%, alongside YouTube, Reddit, CNBC, and LinkedIn (source: wiki/brandlight-where-ai-gets-answers-2025.md). Four platforms, four completely different citation universes — measured across 50 million user journeys.
The overlap between those universes? Less than 1% between ChatGPT and Perplexity citations (source: wiki/ai-ranking-factors-brandfeatured.md). The platforms your customers use most are drawing from source pools that are, for practical purposes, non-overlapping.
The Local Business Case Is the Starkest Illustration
Local Falcon’s own research across 190,000 ChatGPT results found that 83% of restaurants are completely invisible on ChatGPT. On Google Search, the equivalent figure is 14% (source: wiki/localfalcon-ai-search-overlap-2026.md). Three years of Google Business Profile optimisation, local citation building, and geographic ranking — almost none of it carries into ChatGPT’s retrieval layer.
That is not a failure of content. It is a failure to understand which index the content needs to be in.
Architecture Also Determines Accuracy
The infrastructure difference is not just about coverage — it determines data quality too. SOCi’s 2026 Local Visibility Index found that Gemini returns accurate business profile information 100% of the time. ChatGPT is accurate 68% of the time, meaning it is wrong about a business’s name, address, phone, or hours nearly one-third of the time (source: wiki/localfalcon-ai-search-overlap-2026.md). Gemini draws directly from Google’s verified business data. ChatGPT doesn’t have that feed. The result is systematic misinformation, at scale, about brands that have no idea it’s happening.
Investment Priority Follows the Architecture Map
The practical implication is simple but almost universally ignored in AI visibility programmes: the platform your customers use most should drive where you invest, and investment in one platform does not substitute for another.
If your customers are using ChatGPT for research and purchase decisions, your Google ranking is largely irrelevant to your AI visibility. The Wikipedia representation question, the earned media coverage in sources ChatGPT actually cites, the question of whether your brand is accurately modelled in ChatGPT’s retrieval layer — these are distinct workstreams from anything Google-related. Conversely, if your customers predominantly use Google, the 86% AIO overlap means traditional SEO is already doing most of the work, and marginal investment in ChatGPT-specific strategies may have very poor returns.
The arXiv study adds one more layer: for well-known brands, AI rankings are largely stable regardless of fresh content, because pre-training priors dominate. For mid-market and niche businesses — where AI models lack strong internal knowledge — retrieved content actively changes what gets cited, query by query (source: wiki/arxiv-web-search-vs-genai-2601-16858.md). Retrieval architecture and pre-training architecture interact, and they favour different brands at different scales.
“Optimise for AI” as a category will age about as well as “be on social media” did. The brands that win AI visibility are the ones that get specific — about which platform, which retrieval architecture, and which source types that platform is actually indexing. The rest are producing content for an index that isn’t reading it.
AI search engines don’t rank pages the way Google does. They decide, page by page, whether a piece of content is worth quoting in an answer — and the data now shows that decision comes down to roughly the same six things, again and again, across independent studies.
Revised: 2026-09-24
The Short Version
Multiple studies — practitioner datasets covering thousands of campaigns, and academic research — converge on the same handful of factors that predict whether AI systems cite a business. None of them require a large budget. All of them require consistency.
The Six Factors
1. Can AI actually read the page? This comes before everything else. Adobe’s analysis of over a trillion US retail site visits found that product pages are, on average, only 66% readable to AI systems — homepages fare a little better at 75%. The most common cause is JavaScript-rendered content that never reaches the raw HTML an AI crawler sees. A page that fails here doesn’t get a lower score on the other five factors. It gets ignored entirely.
2. Does the content answer the question immediately? Content that states its main point in the first sentence of each section — rather than building up to it — is cited roughly 2.3 times more often than content with a traditional narrative structure, according to campaign data spanning hundreds of clients. AI systems extract answers; they don’t reward you for making them scroll to find one.
3. Is it written plainly? Content that’s needlessly hedged, padded, or complex costs citations even when the underlying facts are accurate. An analysis of how ChatGPT selects and quotes source material found that direct, definitional sentences — “X is Y” — are cited nearly twice as often as explanatory or discursive prose, and that headings framed as direct questions are twice as likely to be quoted as narrative ones. AI systems extract whichever available source states a fact most clearly — not necessarily the most authoritative-sounding one.
4. Is it current— or at least maintained? AI-cited content skews newer than traditionally-ranked content. University of Toronto researchers measured the typical age of the pages each system drew on: in consumer electronics, 62 days for Claude and 80 for ChatGPT, against Google’s 130; in cars, 148 and 162 against Google’s 493. But the more useful finding is that recency is mostly a maintenance job, not a publishing one. Across roughly 47,000 citations tracked between March and June 2026, three-quarters of the pages AI tools drew on had been updated within the previous year — yet fewer than half had been published in that window, and more than a quarter of the pages counting as fresh first went live over two years ago. Updating an existing page counts — we covered this in detail in Updating Your Best Pages Beats Publishing New Ones.
5. Are you mentioned by name, not just linked? Across 75,000 brands, Ahrefs found that mentions of a brand across the web tracked most closely with its visibility in Google’s AI Overviews, at 0.664 — against 0.218 for backlinks, the currency of search optimisation for two decades. That is an association rather than a proven cause, as Ahrefs itself stresses. But the direction is hard to miss: a link is a reference, a named mention is a recommendation.
6. Is your business unambiguous? Name, address, and category consistency, plus basic structured data markup, form a disambiguation layer underneath everything else. Without it, the other five factors get diluted across fragmented, unrecognised versions of your brand — the AI can’t credit consistent evidence to an entity it can’t reliably identify as one thing.
What This Means in Practice
None of these six factors require a large marketing budget. They require structural discipline: checking whether your pages are actually machine-readable, writing answers before context, keeping your most important pages current, and making sure your brand is described the same way everywhere it appears. Most businesses have never audited even the first one.
In May 2026, Google published a guide to generative AI optimisation. The headline message: “Optimising for generative AI is still SEO.” Nick Fox, Google’s VP of Search, has said the same thing in public. A senior Google executive, using the authority of one of the most trusted technical brands on earth, telling marketers clearly: the skills you have are the skills you need.
The statement is correct. It is also written by Google, for Google. And Google is no longer the only AI your customers are using.
What the data actually shows
The arXiv study that tested this most rigorously (Chen et al., University of Toronto, January 2026) measured domain-level overlap between each major AI system and Google’s top-10 search results across 1,000 queries:
GPT-4o: 4.0% overlap with Google
Gemini: 11.1%
Claude: 12.6%
Perplexity: 15.2%
GPT-4o — the engine behind the most-used AI chatbot on the planet — shares four percent of its cited domains with Google’s top results. Not forty percent. Four.
BrandFeatured’s analysis adds a further dimension: less than 1% of citations overlap between ChatGPT and Perplexity. The two most-used AI answer engines are drawing from almost entirely different source pools.
Google’s own AI products sit at the high end of the overlap range (Gemini at 11%). This makes sense: Gemini is built on Google’s infrastructure and uses Google Search grounding. Of course it draws from Google’s index. That’s also why Google’s guide is correct — for Google’s AI.
The guide answers the right question about the wrong platform
Nick Fox’s statement is accurate. For Google AI Overviews, 76% of citations come from pages already in Google’s top-10 results (Ahrefs data). Rank in Google, appear in Google’s AI. The logic holds. The advice works.
The problem is the implicit premise: that optimising for Google is optimising for AI search. That premise was once defensible. A strong Google presence got you most of the way there across platforms.
By mid-2026, that premise has collapsed. A page that ranks well in Google has roughly a one-in-twenty chance of appearing in GPT-4o’s cited sources for the same query.
Google’s guide doesn’t mention this. It doesn’t need to. It is a guide to Google’s products.
The platforms diverge in ways that change the strategy, not just the tactics
The divergence is not just about which pages each AI retrieves. It runs through every lever brands can actually pull.
Earned media vs owned content. The arXiv study found GPT-4o cites 57% earned media; Gemini cites 46% earned and 46% brand-owned content. Google’s own AI gives substantially more weight to brand-owned pages than ChatGPT does. Optimising your website for Google’s AI is rational. That same owned content is a marginal signal for ChatGPT, which is dominated by independent editorial coverage.
Freshness windows. ChatGPT’s median citation age in consumer electronics is 80 days. Google’s is 130 days. A content programme calibrated to Google’s freshness cadence — one that refreshes every quarter — may be running too slow to reach ChatGPT’s retrieval window for the same queries.
Outlet profile. Claude cites Reuters roughly 50 times less than ChatGPT. Claude’s top-cited journalism outlets include Good Housekeeping, TechRadar, and Harvard Business Review. ChatGPT and Gemini share an almost identical outlet profile anchored by Reuters, the Financial Times, Time, Forbes, and Axios. A PR programme built for Google’s citation preferences will serve Gemini well and largely miss Claude.
These are not marginal differences. They are structural divergences across the platforms that now collectively account for the majority of AI-mediated information discovery.
The market share shift makes this urgent
ChatGPT’s share of AI chatbot traffic was 87% in early 2025. By May 2026 it had fallen to 64%. US mobile share had dropped below 40% for the first time. Gemini went from 5.7% to 21.5% in the same period.
Most AI visibility advice was written for an 87%-ChatGPT world. That world is gone. The distribution across AI platforms is fragmenting rapidly — which means the gap between “optimise for Google” and “optimise for AI” is widening every quarter.
Google’s guide was accurate when it was published. The market it describes is changing underneath it.
What Google’s guide gets right (and why that still matters)
The guide’s underlying recommendations are not wrong. Demonstrate expertise, publish original research, earn independent editorial coverage, maintain consistent entity signals. These are correct — and they are the inputs that help with ChatGPT, Perplexity, and Claude as well as Google’s AI.
The fundamentals converge. The execution diverges.
Which outlets you target for earned coverage. How fast you refresh content. Which platforms you prioritise when budgets are constrained. Those decisions are platform-specific. Google’s guide does not help you make them, because it was not written to.
The question the guide doesn’t ask
Google’s guide starts from the premise that you are optimising for Google. That is a reasonable premise for a guide published by Google.
It is not a sufficient premise for a brand strategy in 2026.
Your customers are not on one platform. A B2B buyer researching software options may be using Claude or Perplexity. A consumer comparing products may be in ChatGPT on iOS. The AI-mediated moments that shape purchase decisions are distributed across platforms with fundamentally different citation logic.
The question the guide doesn’t ask is the one that matters most: which AI is your customer using when they form the opinion that drives the decision?
Google’s guide is the right answer. It’s just the answer to a narrower question than the one you need to be asking.
Sources: Google AI Optimization Guide (https://developers.google.com/search/docs/fundamentals/ai-optimization-guide, 2026-05-15); Chen et al., arXiv 2601.16858 (University of Toronto, January 2026); BrandFeatured AI ranking factors analysis (2026); Ahrefs AI Overview citation study; Muck Rack Generative Pulse (via Nieman Lab, July 2025).
Why AI Search Engines Default to Famous Brands (and Why That Doesn’t Last)
AI search engines recommend famous brands more often not because those brands are better, but because they were already well documented before the AI was ever built. That’s a fixable gap, not a permanent one — and for most of the questions real customers actually ask, it doesn’t apply at all.
The Short Version
Every AI system — ChatGPT, Gemini, Google’s AI Overviews — was trained on a huge snapshot of the internet before it ever answered a single question. Big, famous brands had spent decades building up Wikipedia pages, news coverage, and mentions across the web. That presence got baked into the AI during training. Small and mid-sized businesses, however good they are, usually hadn’t built up the same paper trail — so the AI never learned much about them.
That’s the actual mechanism behind “AI only recommends famous companies.” It isn’t a judgement of quality. It’s a gap in what the AI happened to read before it was finished being built.
The Evidence
A 2026 University of Toronto study tested this directly. Researchers fed an AI model manipulated evidence about different brands and measured how much its answers changed. For famous brands, changing the evidence barely moved the AI’s recommendations at all — its mind was already made up from training. For lesser-known brands, the same manipulation changed the results dramatically. The AI had no fixed opinion to defend, so whatever it found when it looked things up is what actually decided the answer.
In other words: fame makes an AI’s opinion sticky. Obscurity makes it responsive.
Where Small Businesses Actually Compete
That second finding is the opportunity. Most real customer questions — “best plumber near me,” “who’s reliable for X,” “is this company any good for Y” — aren’t the kind of query where an AI has a fixed opinion already. For questions like these, the AI has to go and look at current information to answer. That’s exactly the territory where a small business’s recent reviews, press mentions, and clear web presence can shape the answer directly.
The famous-brand advantage is real, but it’s concentrated in the broad, generic searches that were already decided years ago. It doesn’t extend to the specific, current, local questions that make up most of what customers actually type.
What This Means in Practice
Beating a famous competitor’s name recognition isn’t the goal, and isn’t realistic. The goal is making sure that when an AI system goes looking for current evidence — which it does constantly, for exactly the questions where small businesses compete — it finds consistent, credible information about your business. That’s a matter of current evidence, not decades of accumulated fame.
Go Deeper
This piece summarises the underlying research and its philosophical framing — including the specific study data and the academic argument for why this amounts to more than a marketing disadvantage — in Art #0049: The Structural Exclusion Problem.
E-E-A-T Is Entity Confidence — So Why Don’t We Say That?
Summary: A provocation piece arguing that Google’s E-E-A-T framework and the practitioner concept of “entity confidence” describe identical phenomena — and examining what is genuinely new about the AI measurement layer.
Last updated: 2026-07-07
Type: Original article / insight post
The GEO industry has spent two years building new vocabulary. Entity confidence. Citation frequency rate. AI visibility scores. Semantic authority. Confidence language analysis. These terms fill product decks, vendor websites, and optimisation guides published from 2024 onwards.
Google invented most of this framework in 2014 and called it something different.
But there is a catch — and it matters more every month. E-E-A-T was designed for one search engine. Half the search behaviour it was built for is now happening somewhere else entirely.
What E-E-A-T Measures
E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness — is Google’s formal quality framework. Google’s HJ Kim described it as “a template they use to rate every single site for every single query.” Marie Haynes distilled the core of it more precisely: “E-E-A-T is a measure of the legitimacy of your entity as a destination for the topics you cover.”
The measurement mechanism was described by Gary Illyes at Pubcon 2018: “E-A-T is largely based on links and mentions on authoritative sites. If the Washington Post mentions you, that’s good.”
Danny Sullivan made the ranking connection explicit: E-E-A-T is not a directly disclosed score but a cluster of proxy signals — third-party mentions, backlinks, review reputation, entity recognition — that approximate what human quality raters would assess.
And the loop closes formally through Pandu Nayak’s antitrust testimony: quality rater assessments generate the Information Satisfaction (IS) Score, and IS-scored documents train the deep learning systems that power Google Search. E-E-A-T signals → rater feedback → IS Score → ranking model training. It is not theoretical; it is a documented production mechanism.
What Entity Confidence Measures
Entity confidence score — as described by the vendors building AI citation measurement tools — is a composite measure of how certain an AI model is about the accuracy and authority of information associated with a brand. Primary vendors measure it through: citation frequency (how often AI mentions a brand for relevant queries), confidence language (whether AI uses “according to Brand X” versus “some sources suggest”), and response position (whether the brand appears first or sixth in an AI answer).
The signals that build a high entity confidence score: authoritative off-site mentions, structured entity data (schema, sameAs markup), cross-platform consistency, third-party recognition from credible sources.
The Mapping
Hold the two frameworks side by side:
E-E-A-T Component
Entity Confidence Equivalent
Authoritativeness — cited by credible third parties
Relationship Mapping — industry ecosystem recognition
The signal overlap is not approximate — it is complete. E-E-A-T is built from off-site mentions, links, entity recognition, review reputation, schema, and community presence. Entity confidence is built from the same list.
The data confirms this at the empirical level. Ahrefs analysed 75,000 brands against AI Overview citation outcomes. The strongest correlate with AI citation was branded web mentions (off-site) at a correlation coefficient of 0.664. Not schema. Not content length. Not author bios. Earned media mentions — the Gary Illyes signal, showing up again in 2025 data for an entirely different platform.
The finding replicates across independent datasets. 82% of AI citations across 1M+ analysed prompts came from earned media (Muck Rack/MacroLingo). And 76% of AI Overview citations came from pages already in the top-10 search results — meaning AI answers are substantially inheriting Google’s E-E-A-T rankings rather than computing something new from scratch.
Did AI Engines Inherit E-E-A-T — or Reinvent It?
The most important question the data raises is whether AI citation systems independently arrived at the same signal set as Google, or simply inherited it.
For Google’s own AI surface, the evidence points clearly to inheritance. In the Ahrefs 1.4M-prompt study, 88% of ChatGPT citations came from the search channel — pages indexed and ranked in traditional web search. For Google AI Overviews and AI Mode, this is even more direct: those systems are grounded against Google’s index by design. AI Overviews are, to a significant degree, Google ranking with a synthesised output layer.
The chain is documented end to end for this surface:
E-E-A-T signals → off-site mentions and links → IS Score trains ranking models → pages rank in search → AI retrieval pulls from ranked pages → AI citation emerges
There is no layer in that chain where AI engines are separately deciding what constitutes a trustworthy source. They are, in large part, delegating that judgment to Google’s ranking systems — which were themselves trained on E-E-A-T rater feedback.
But this chain only describes part of what is now happening.
The Search That Google Doesn’t See
In June 2025, OpenAI published its first major usage research. The finding that matters here: 51.6% of all ChatGPT interactions are now search-like information queries — the primary use case for the platform, having overtaken content generation in a single year. ChatGPT has 700 million weekly active users. At 11–12% monthly growth from early 2025, it passed the mass adoption threshold around April 2026.
These are not Google searches with a different interface. They are queries going directly to an AI engine that is not Google — and they are not passing through Google’s ranking systems before generating a response.
The University of Toronto’s peer-reviewed study of 1,000+ queries (Chen et al., arXiv:2601.16858, January 2026) makes the structural difference measurable. Domain-level overlap between GPT-4o’s citations and Google’s top-10 results: 4%. Claude’s overlap: 12.6%. Perplexity’s: 15.2%.
GPT-4o is making its own authority decisions on 96% of the domains it cites. It is not routing those decisions through Google’s ranking systems. For that 96%, E-E-A-T → search ranking → AI citation is not the chain. The AI is drawing on training data, its own retrieval logic, and entity signals that exist independently of Google’s index.
The pre-training bias finding from the same study sharpens this further. For well-known brands, AI rankings remain highly stable even when retrieved supporting evidence is removed or shuffled entirely — the model’s pre-trained understanding of brand authority dominates regardless of what is currently ranking on Google. For those queries, Google’s ranking is not the input. Training data is. And training data is shaped by years of accumulated web content, entity signals, and third-party coverage — not by last week’s ranking positions.
Zero-click search has been driving this shift at the structural level: approaching 65–70% of queries now resolve in AI-generated answers without a click-through. Google AI Mode runs at approximately 93% no-click. The search that is happening is increasingly not search in the traditional sense — it is AI-mediated information retrieval across multiple platforms, only some of which route through Google.
Then What Is Entity Confidence Actually Adding?
If E-E-A-T is entity confidence, the concept is not new. But what it covers — and how it is measured — genuinely is.
E-E-A-T was built to track brand authority in one system: Google Search. The signals are right. The gap is the scope.
Entity confidence, properly defined, covers two surfaces that E-E-A-T tracks only partially:
The Google-adjacent surface: AI Overviews, Google AI Mode, and AI systems that retrieve primarily through search indices. Here, E-E-A-T signals are the right inputs and search ranking is a reasonable proxy for AI citation eligibility. This is the surface where the inheritance chain holds.
The direct AI surface: ChatGPT queries without web search enabled, Claude, Perplexity, and all AI interactions where the model answers from training knowledge or its own retrieval logic rather than Google’s index. Here, E-E-A-T signals still matter — the underlying signals (earned media, entity clarity, cross-platform consistency) are what train these models and what their retrieval systems learn to trust. But Google ranking is not a reliable proxy for AI citation on this surface. A brand can hold Google position 3 and have no meaningful AI presence if its entity signals are thin in the training corpus.
This is the surface that is growing fastest. The 51.6% of ChatGPT interactions that are search-like queries is not a static number — it is a rising proportion of a user base that grew 75% in five months.
The second genuine addition is the measurement layer. E-E-A-T has no disclosed score. It is tracked by proxy: rank improvement, traffic, domain authority. These are position-based, deterministic metrics.
AI citation is stochastic. Only 30% of brands maintain consistent visibility across multiple regenerations of the same query. A brand is not in AI position 3. It appears in approximately 65 out of every 100 relevant AI responses. That is a probability distribution, not a rank — and it can differ significantly across AI platforms using different retrieval architectures and training histories.
The third addition is the narrative framing layer. Google returns a ranked list. AI platforms synthesise a recommendation and present it as a direct answer. There is a meaningful difference between being described as “the industry standard” and “a budget-friendly alternative” — both are appearances, but at very different confidence levels. That qualifier is invisible in any rank tracking tool.
What This Means for Strategy
Brands being sold AI visibility programmes as something new and distinct from their existing SEO and PR investment should ask: what specifically are we being asked to do that we are not already doing?
For the Google-adjacent AI surface, the honest answer is: at the foundational level, very little. The inputs are shared:
For the direct AI surface — the faster-growing, non-Google half — the inputs are still the same signals. But the tracking is not. Google rankings tell you nothing about how Claude or ChatGPT without web search represents your brand. Measuring that requires different instruments: repeated sampling across regenerated prompts, platform-specific citation profiling, confidence language analysis. This is what EC measurement adds that rank tracking cannot.
The framing shift matters too. Google Search is where you compete for rank position. Direct AI search is where AI decides whether to recommend you based on what it has learned about your brand — and that learning happened before the user typed a word.
The Vocabulary Question
Why doesn’t the entity confidence industry say “we’re measuring E-E-A-T for AI engines”?
The uncharitable read: novelty sells. New vocabulary justifies new products and new budget lines.
The more generous read — and the more accurate one — is that E-E-A-T is a Google framework, and the problem is now multi-platform in a way it was not when the GEO vocabulary was being built. Saying “build your E-E-A-T” implies Google Search is the destination. Saying “build your entity confidence” points to the full citation surface: Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini — including the surfaces that have no relationship to Google’s ranking systems.
That vocabulary change reflects a real scope change. The underlying signals — what you need to do to build brand authority — are E-E-A-T. The measurement layer, the platform coverage, and the probabilistic framing of visibility scores are new. And the urgency is new: a search behaviour that took Google a decade to accumulate is now being replicated across multiple AI platforms at mass-adoption pace.
The question worth asking is not whether your E-E-A-T programme is doing its job. It is whether you know what your brand looks like on the surfaces that E-E-A-T tracking was never designed to see.
When a journalist writes something inaccurate about your business, you know who to call. There is an author, an editor, a publication. The correction path is clear, even if it’s difficult.
When AI says something inaccurate about your business — wrong founding date, conflated with a competitor, fabricated capabilities presented as fact — there is no one to call.
This isn’t a gap in current infrastructure that will be filled as the industry matures. It is a structural feature of how AI-generated content works, and Harvard’s Kennedy School has a name for it: distributed agency.
No one to call
The Harvard Kennedy School Misinformation Review draws a sharp distinction between two fundamentally different categories of inaccuracy.
Human misinformation is produced by actors with beliefs, motivations, and goals. The correction strategies follow from that: identify the actor, apply social or legal pressure, issue counter-statements, reduce amplification. The accountability chain exists because there is an agent at the origin.
AI hallucination emerges from a different structure entirely. A probabilistic system generates statistically likely text sequences — predicting the next token based on patterns across billions of documents. There is no intent. There is no understanding of accuracy. There is no actor who made a decision about your brand. The output is the aggregate product of training data patterns meeting a query.
Grok 3 sent 154 of 200 users to error pages in the CJR/Tow Center study. It did so with no awareness that anything had gone wrong. ChatGPT correctly identified one of ten San Francisco Chronicle articles despite Hearst’s formal content partnership with OpenAI. OpenAI did not decide to misrepresent those articles; the model generated plausible text and the text was wrong.
The practical consequence is that you cannot send a correction to a probability distribution. The remediation path for AI-generated brand misinformation is fundamentally different from correcting a journalist, a review, or a social media post. There is no actor to change their mind.
Why the belief doesn’t update
If there’s no actor to correct, you might assume the fix is at least clean: insert accurate information into the training and retrieval layers and let the model update. The reality is harder.
Mark Coeckelbergh (University of Vienna, Social Epistemology, 2025) describes the structural dynamics of AI-mediated belief as an “economy of belief revision.” Maintaining an existing belief is the cheaper option — it requires less work. Revising a belief requires more. The architecture is structurally biased toward its current representation of the world. “Using AI is believing.”
The parallel for AI systems themselves is direct. A model that has formed a representation of your brand — its authority tier, its associations, its positioning — holds that representation with inertia. Training-mode knowledge, once embedded, is not updated by a single counter-publication. It is shifted by sustained, consistent signals across multiple retrieved sources over time.
Coeckelbergh’s broader concept sharpens this further. AI-generated brand claims are not testimony in the ordinary sense — there is no human originator who holds the belief that needs revising, who can be questioned, who can update their view. They are a new epistemic category: machine-generated assertions delivered in natural language, without a human at the origin, resistant to the correction mechanisms we normally apply because those mechanisms assume an agent.
This is why correcting AI brand representation is not a one-shot content operation. It requires sustained provenance infrastructure — consistent signals across earned media, structured data, and knowledge graph entries — built and maintained across retrieval cycles to overcome structural inertia. One press release does not move the needle. One corrected Wikidata entry does not move the needle. The pattern of evidence, accumulated over time, is what shifts the model’s representation.
From answering to acting
The distributed agency problem would be serious if AI only answered questions. It becomes structurally different as AI moves from information to action.
The spectrum of agency runs from direct human browsing at one end to fully autonomous AI action at the other. The middle position — AI-assisted human activity, where software executes tasks on behalf of humans who have stated a goal — has been steadily moving toward the autonomous end.
Large Action Models are the technical substrate making this concrete. Where Large Language Models answer questions, LAMs take actions: executing task sequences, navigating interfaces, filling forms, making decisions in dynamic environments. OpenAI’s Operator and Google’s Auto Browse in Chrome can already compare products, fill forms, and make purchases without human intervention. Baidu’s production AI search system runs a four-agent architecture — a master agent classifying query complexity, a planner decomposing it into sub-tasks, an executor running tools, a writer synthesising the output — all without a human in the loop.
When AI answers a question about your brand inaccurately, a potential customer may be misled. The harm is probabilistic, downstream, mediated by human judgment.
When AI acts on behalf of a user — selecting vendors, booking suppliers, making purchase decisions based on whatever knowledge it has assembled — wrong brand knowledge doesn’t mislead. It determines concrete commercial outcomes, without any human review step between the AI’s representation and the result.
The accountability gap does not get smaller as AI becomes more capable. It gets wider, because the consequences of distributed agency are no longer hypothetical.
The trust chain
Gartner identifies trust as the primary adoption constraint for agentic AI: users will only delegate real decisions to agents they trust. But trust works in a chain.
If users trust agents, those agents must in turn trust the sources they consult. An agent making a vendor recommendation is implicitly asserting the credibility of whatever information it has assembled about that vendor. If that information is wrong — stale, conflated, fabricated — the trust chain carries the error forward, at scale, automatically.
Coeckelbergh’s concept of epistemic bubble is directly relevant here. Once a brand is positioned within a particular cluster in an AI system’s training or retrieval logic, counter-evidence may be de-weighted by the same mechanisms that created the cluster. AI-mediated knowledge environments can sustain incorrect representations not through malice but through the structural logic of how relevance and consistency are weighted. A brand that has been miscategorised, or weakly evidenced, or consistently underdescribed relative to competitors, faces an uphill correction problem that gets steeper the longer the representation persists.
The brand that is well-represented in the AI’s knowledge layer — consistent entity signals, authoritative third-party coverage, structured data that is unambiguous — sits inside the trust chain. The brand that is poorly represented, misidentified, or absent sits outside it.
When agents move from recommendation to action, being outside that trust chain is not a visibility problem. It is a market access problem.
The only available lever
None of this changes by filing a complaint. The accountability is distributed; the responses must be structural.
The Harvard HKS framework is explicit about this: the remediation path for distributed agency problems runs through the supply side. What the model has learned, and what it can retrieve, are the inputs that determine what it says and does. Those inputs are the only levers available to brands.
In practice, this means:
Earned media in publications the model retrieves reliably. The 82% of AI citations that come from editorial sources (Muck Rack, 1M+ citations) are the primary signal pool. The model learns which brands belong in which categories largely from this layer.
Structured entity data — schema markup, sameAs links, Wikidata entries — that gives the model unambiguous resolution of who you are and what you do. The technical layer that prevents conflation and suppresses drift.
Consistent, sustained signals across retrieval cycles. Not one-shot publications, but ongoing patterns of corroboration. The economy of belief revision works against you on one input; it works for you when the signals are consistent enough to shift the prior.
The question “who is responsible for what AI says about your brand?” has a clean answer: no one who can be corrected.
The question that follows is different: what is in the supply layer that shapes what the model has learned?
That is what entity confidence measures — not accountability for the current representation, but the reliability of the evidence available to AI systems when they reach for your brand. As AI moves from answering to acting, the stakes of that measurement become concrete in a way they weren’t when the only output was text.
A weak entity signal was a visibility problem when AI gave advice. When AI makes decisions, it becomes something closer to market exclusion.
Entity Confidence measures the reliability of that evidence layer.
When businesses begin addressing their AI visibility, they face a fundamental strategic question: should they focus on quick fixes that produce immediate improvement, or invest in deeper changes that build sustainable competitive advantage? The most effective approaches typically combine both, but understanding the distinction helps set appropriate expectations.
The Quick Win Opportunity
Many businesses have low-hanging fruit—problems that can be fixed quickly with relatively little effort. Inconsistent information across platforms is often the prime example. If your business name is slightly different on three directories, your old phone number appears on two review sites, and your address format varies everywhere, fixing these issues takes hours rather than months.
These quick wins matter because they remove obstacles to AI confidence. They don’t build new strengths, but they eliminate weaknesses that may be preventing recommendations. For businesses with substantial inconsistency issues, this cleanup can produce noticeable improvement relatively quickly.
Other quick wins might include claiming unclaimed business profiles, completing sparse listings with richer information, or ensuring your Google Business Profile is fully optimised. These actions have outsized impact relative to the effort required.
The Strategic Investment Reality
Genuine competitive advantage in AI visibility comes from factors that take time to build. Developing a substantial review presence doesn’t happen overnight. Building thought leadership that attracts media coverage requires sustained effort. Accumulating professional accreditations and industry recognition takes years in some cases.
These strategic investments are harder to execute but also harder for competitors to replicate. While anyone can clean up their directory listings, not everyone can build genuine industry expertise and recognition. The businesses that invest in these deeper factors develop moats around their AI visibility position.
The Phased Approach
Sensible improvement strategies typically proceed in phases. The first phase addresses obvious issues: inconsistencies, incomplete profiles, basic optimisation gaps. This creates a clean foundation and often produces encouraging early results.
Subsequent phases build on this foundation, pursuing improvements that require more time and effort. Perhaps developing a systematic review generation programme, creating substantive content that demonstrates expertise, or pursuing relevant professional recognition. Each phase builds on the previous, creating cumulative improvement.
The Ongoing Nature of the Work
Unlike a website redesign—which has a clear beginning, middle, and end—AI visibility improvement is ongoing work. The landscape continues to evolve. New AI systems emerge with potentially different evaluation criteria. Competitors improve their positions. Your own business changes in ways that need to be reflected in your digital presence.
This ongoing nature has implications for how businesses should think about investment. Rather than a project with a fixed budget and timeline, it’s more like marketing or customer service—a sustained operational function that requires ongoing attention and resources.
The Competitive Dimension
All of this happens in a competitive context. Your visibility is assessed relative to alternatives. If competitors are improving their AI presence while you focus solely on quick fixes, your relative position may decline even as your absolute position improves. Strategic investment isn’t just about building strength—it’s about maintaining advantage as the field evolves.
The right balance of quick wins and strategic investment depends on your starting position, competitive context, and business objectives. Understanding what improvements are available at each level enables informed planning and realistic expectations.
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behavior or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
