AI SEO statistics

Why this hub is different

A roundup that treats every number with equal confidence is a real problem. It's the exact failure the provenance audit documents across this field. This hub applies that audit's grading to every statistic published on this site. It's all in one scannable table.

The grade column is the whole product. Without it, this is a list of links. With it, you can tell at a glance which figures will survive a client asking where the number came from, and which will not.

The master table

How to read the grade column

Traceable means the study behind it shares its sample, method, and date. Partial means a named source stands behind it, but the full method isn't public. Broken chain means the claim itself doesn't line up across the sources we checked. See the linked page for the specifics in each case.

How a grade actually gets assigned

The grades aren't impressions. Each one comes from running the same five checks against the underlying source, the same procedure used in the provenance audit.

Check one: is there a named source at all? Not "studies show." A specific organisation, a specific publication. A figure that fails here is graded broken immediately, regardless of how plausible it sounds.

Check two: does the link lead to original data, or to another article? This is where most figures fail. A citation chain that runs blog to blog to blog and never reaches a dataset is a figure with no foundation, however many times it has been repeated.

Check three: is the sample stated? How many queries, how many domains, over what window. A percentage with no denominator cannot be checked or reproduced by anyone.

Check four: is the method described well enough to repeat? This is the line between partial and traceable. A vendor can name itself, state a sample size, and still not describe how the measurement worked. That earns partial, not traceable.

Check five: does it reconcile with other figures on the same thing? Where two credible sources report incompatible numbers for the same quantity, and the difference cannot be explained by different populations, the claim gets graded broken until someone reconciles them.

A figure that passes all five earns Traceable . Passing one through three but not four earns Partial . Failing check one, two, or five earns Broken chain . That is the whole system, and it is deliberately simple enough to apply consistently.

What each page actually found

A table of links tells you where to go. It doesn't tell you what you'd find there. Here is the headline finding from each linked page, with the reason it carries the grade it does.

AI Overviews. The share of AI Overview citations coming from Google's top-10 results fell from roughly 76% to roughly 38% year over year, per Ahrefs. Graded traceable because the same organisation ran both measurements with a comparable method and disclosed sample sizes both times. That makes it one of the few genuine trend claims in the field rather than two unrelated snapshots.

AI Mode and query fan-out. Google's own patent documents eight sub-query type categories. The widely repeated "8-16 sub-queries per question" figure could not be traced to any disclosed-method study. Graded partial: the mechanism is documented, the specific count is folklore.

ChatGPT. Citations skew heavily toward Wikipedia and encyclopedic sources. Google ranking predicts ChatGPT citation weakly, around 12% overlap, far below AI Overviews. The ranking-overlap figure traces to Ahrefs with a disclosed sample. The source-skew figure is vendor-reported from a non-public corpus, which is why the page carries partial.

Claude. The least-measured major surface. No large-scale citation study exists publicly. The page is deliberately shorter than its siblings, and that absence is itself the finding rather than a gap in our research.

Perplexity. Citations skew heavily toward Reddit. Ranking overlap sits around 33%, notably higher than ChatGPT's, making it the engine where classic SEO transfers most directly. The overlap figure is traceable; the Reddit-skew figure is vendor-reported.

Gemini. Graded broken for a structural reason rather than a sourcing one. Gemini, AI Overviews and AI Mode are three distinct Google surfaces that coverage routinely conflates, so figures attributed to "Gemini" frequently describe something else.

Cross-platform domain skew. Each engine has a distinct centre of gravity: encyclopedic for ChatGPT, community for Perplexity, multimodal for AI Overviews. All figures come from one vendor corpus that isn't independently public. Partial throughout.

Crawler economics. AI crawlers take dramatically more pages per referral than classic search crawlers do, spanning three orders of magnitude between the most and least extractive. Cloudflare-derived, reproduced via secondary aggregators rather than pulled directly, hence partial.

Market share. AI referrals remain around 1% of the average site's traffic while growing fast, and ChatGPT's share of the AI category fell as competitors gained. The page's real contribution is the denominator warning: a within-category percentage quoted as a whole-traffic figure is the most common error here.

Zero-click. Graded traceable, and the most useful disambiguation page on the site. Three separate legitimate studies produce numbers near 58% that measure entirely different things. The figures are fine. Collapsing them into one is the problem.

Referral traffic. AI platforms send small but growing traffic volumes, and standard analytics likely undercount it because much AI-sourced reading never produces a click at all. Partial.

Conversion. Graded broken, and deliberately so. Published claims range from roughly 4.4x to 23x organic conversion. That is not a range, it is different orders of magnitude measuring different things, and no source reconciles them.

GEO tactics. Graded per-tactic rather than as a page. Four of fourteen tracked tactics carry fact-grade evidence, all from the single peer-reviewed study. Most of the rest are widely recommended and untested.

Commerce. Reported growth in AI-driven store traffic is real and substantial, from what was likely a very small base. The much-quoted forecast about agent-influenced transactions is a projection, not a measurement.

Local search. Graded broken as a structural gap. Local intent is a large share of all search behaviour and rigorous AI-search research on it is almost entirely absent. The page states that plainly instead of borrowing a general-purpose number.

Concordance. A reported ~11% domain overlap between ChatGPT and Perplexity. Vendor-reported, two engines only. If it holds up, it is the strongest argument in the field for measuring each engine separately.

The strongest figures in the field

Three figures on this site earn traceable grades, and it is worth understanding what they have in common, because it describes what good measurement looks like here.

The AI Overview ranking-overlap trend. Same organisation, two points in time, comparable method, disclosed sample sizes. That combination is what allows a genuine trend claim rather than two disconnected snapshots. Most "X changed to Y" claims in this field fail precisely because the two numbers came from different methods.

The Princeton GEO study's effect sizes. Peer-reviewed, disclosed method, a stated benchmark of roughly 10,000 queries, and a built-in negative control that actually came back negative. That last detail matters more than most readers notice: a study that finds an effect everywhere it looks has not demonstrated it can detect a failure.

The SparkToro zero-click figure. A named clickstream panel, a stated geography, a stated date. It is frequently misused, but the underlying measurement is sound. The failure is in how people quote it, not in how it was produced.

Notice none of these are the most exciting numbers in the field. The exciting ones tend to be the vendor-reported ones. That inverse relationship, between how quotable a figure is and how well it is sourced, is worth internalising.

The weakest figures, and why

Three categories carry broken grades on this site, for three genuinely different reasons.

Conversion benchmarks: irreconcilable spread. Claims ranging from 4.4x to 23x cannot all describe the same quantity. Some measure session-level conversion, some measure a modelled revenue effect, some do not say. Until someone publishes definitions alongside the numbers, quoting any single figure from that range is picking one at random and presenting it as consensus.

Gemini and local search: structural gaps. Neither is broken because a specific number failed a check. They are broken because the research needed to produce a reliable number has not been done at all. Publishing a borrowed figure from an adjacent surface would fill the gap dishonestly.

Terminology adoption: no disclosed-method dataset. Every "GEO vs AEO, which is winning" page eventually shows a trends chart. We could not find one with a stated method behind it, so that page states the gap instead of fabricating a chart to fill the section.

Naming these plainly is uncomfortable and it is the point. A statistics hub that grades everything favourably is a marketing asset. One that flags its own weakest categories is a reference.

Five patterns that recur across this whole field

Grading sixteen categories one by one surfaces patterns that no single page shows on its own. These are the recurring failure modes, and recognising them is more durable than memorising any individual figure.

One: a small number of original sources, endlessly re-cited. Trace enough figures and the same handful of names appear underneath nearly all of them. A number that appears in forty articles can still be one measurement. Ubiquity is not corroboration.

Two: the measuring parties have a commercial stake. The organisations best positioned to measure AI citation at scale are visibility vendors whose data is the product they sell. That is not an accusation of dishonesty. It is a structural reason why full methods rarely get published, and why so many rows here carry partial rather than traceable.

Three: populations get dropped in transmission. A figure is measured on a specific, stated subset. Each time it is repeated, the qualifier gets a little shorter. Several steps later it describes everything. This is the single most common way an accurate number becomes a false claim.

Four: correlation gets written up as causation. Almost every finding in this field observes that cited pages tend to have some property. Very few show that adding the property causes citation. The language used to report them rarely preserves that distinction.

Five: nothing gets re-measured. Figures are published once and quoted for years, against products that change every few months. A study that was accurate when run may describe a system that no longer exists. Very little in this field gets a second measurement, which is why the AI Overview trend figure stands out so sharply for having one.

Every grade on this page is ultimately an assessment of how many of those five a given figure has survived.

The anatomy of a typical roundup failure

It helps to name what "unverified roundup" actually looks like. It's easy to wave at as a problem without spelling it out. A typical list bullets a number, say "AI Overviews now appear on 47% of searches." It attaches a link. That link often leads to another blog post, not original data.

Repeat that pattern a hundred times and the roundup reads as thorough. But in the one sense that actually matters, trusting any single number, it's mostly unverifiable. This hub's answer isn't fewer numbers. It's the same numbers, with an honest verification status attached to each one.

A worked example: tracing one bad number

Say a figure claims "68% of searches now end without a click." Sounds precise. Step one: click through to the source. If it leads to another blog post instead of a named study, that's already a red flag. Step two: check the population. Does 68% describe all searches, or just one platform's searches? Step three: check the date. A number from three years ago may no longer describe today's behavior at all.

Run any number through those three steps before repeating it. Most roundups skip all three. This hub tries to do that work once, per statistic, so you don't have to redo it every time you read a number.

Statistics we deliberately do not publish

A statistics hub is defined as much by what it leaves out as by what it includes. Several widely circulated figures are absent here on purpose. Naming them is more useful than silently omitting them.

A single composite "AI visibility score." Every major vendor has one. Every one is unfalsifiable by design, because the weighting between inputs is proprietary. Two scores of 78 from different tools are not comparable, and neither can be checked from outside. A hub whose purpose is comparability cannot contain an incomparable metric.

Bare "AI Overviews cut clicks by X%" claims. Not because the underlying studies are bad, but because the figure is meaningless without its population attached. The same number describes very different things depending on whether it covers all searches or only AI-Overview-triggering ones. The disambiguated version lives on the zero-click page.

Precise-sounding percentages with no named source. A figure like "73.2% of marketers say" fails the first check in the grading system. Suspicious precision is itself a signal: a number stated to one decimal place, from a method that could not possibly support that resolution, usually indicates the precision was invented rather than measured.

Forecasts presented as measurements. Projections about what share of transactions AI agents will influence, or how large AI search will become, are forecasts. They can be reasonable and still are not data. Where this site references one, it is labelled as a projection rather than folded into a statistics table.

Anything from a source we could not identify. Several genuinely interesting figures circulate with no traceable origin at all. They may well be accurate. Without a source there is no way to know, and repeating them would add to exactly the problem this hub exists to document.

How this hub is actually built and maintained

Every row above points to a full page elsewhere on this site. Each one carries its own sourcing, its own method notes, and, where it applies, first-party data from the Citation Index research program.

This hub itself holds no original statistics. It's a navigation and grading layer sitting on top of content that already exists in full. That's exactly why it can stay short without losing depth. The depth is one click away, on every linked page.

How to actually use this page

Don't read this table top to bottom looking for one big takeaway. Use it like an index instead. Find the category you need. Check its grade first. Then click through only if both the grade and the topic matter to your work.

If you're citing a number in your own writing, go one step further. Open the linked page. Confirm the number still matches what's written there before you repeat it. Grades can change as new research comes in.

How to cite a figure from this site

If you are quoting one of these numbers in your own work, a few practices make your claim hold up better than most of what circulates.

Cite the original source, not this site. Where a figure comes from Ahrefs or a peer-reviewed paper, name them. This site is a grading and navigation layer over other people's data for most of these rows. Citing the intermediary when the primary source is available weakens your claim unnecessarily.

Carry the qualifier with the number. "Vendor-reported from a non-public corpus" is four words. Dropping them turns a partial-graded figure into something that reads as settled fact, which is exactly how good numbers become bad ones through repetition.

State the population. Every percentage describes some specific set of things. Saying which one is the single most effective habit for not repeating this field's dominant error.

Include the date. These products change fast enough that an undated figure has no shelf life. A number from eighteen months ago may describe a system that no longer exists in that form.

Check the linked page before you publish. Grades and figures both change as new research lands. What was traceable last quarter may have been superseded, and what was broken may have been reconciled.

The biggest measurement gaps, ranked

Reading sixteen graded categories together also makes the holes visible. These are the questions where a reliable number would change how people work, and where none currently exists.

One: does anything a site controls causally change citation rate? The most important open question in the field, and the least answered. Nearly everything published is correlational. Until randomized tests exist, most tactical advice rests on association.

Two: how long does a citation last? Every study is a snapshot. None follow the same citations forward. A citation that vanishes in two weeks and one that persists a year are currently indistinguishable in every published measurement.

Three: do retrieval bots execute JavaScript? Binary, foundational, cheap to test, and still untested publicly. If the answer is no, a whole category of site architecture is excluded from citation regardless of content quality.

Four: what does Claude cite? The least-measured major surface, with no large-scale study at all. Not a hard problem technically. Simply one nobody has done.

Five: how much do engines actually disagree? The ~11% concordance figure covers two engines from one vendor corpus. Whether it holds across more engines, or varies by query type, is unmeasured.

Each of these is registered as a study in the AI Citation Index roadmap, with predictions published before data collection begins. That ordering is the point. A prediction written after seeing results is not a prediction.

Update cadence

This hub updates in substance whenever a linked page changes. Figures live on their own dedicated pages. They are never copied here separately.

That design choice has a specific consequence worth stating. This page cannot go stale relative to its sources, because it holds none of the numbers itself. The common failure of a statistics roundup, where the roundup keeps quoting a figure its source has since revised, is structurally impossible here.

Grades get reviewed whenever new evidence lands on a linked topic. A grade can move in either direction. A partial figure becomes traceable if the underlying vendor publishes its method. A traceable figure becomes partial if a re-measurement uses a method that is not comparable to the original. A broken figure becomes partial or traceable if someone reconciles the conflicting sources.

Movement in the other direction matters too, and it is the direction most reference pages quietly avoid. If a figure that once looked solid stops holding up, the grade drops here and the change is visible. A reference that only ever upgrades its own confidence is not tracking evidence. It is tracking optimism.