A technical reference on AI search visibility. This site sells nothing, takes no engagements and endorses no products. Consulting enquiries are handled separately at hartzer.com.

Hartzer.it.com logoHartzer.it.comAI search visibility reference
The evidence

The AI Search Research Index

The studies this industry quotes at each other, and what each headline number is actually a share of.

An index is a bibliography, not a body of evidence

A research index collects figures from many studies into one list. That is a useful thing to have and it is not the same thing as evidence. The index format strips each number away from the design that produced it, so a figure resting on 115 prompts sits in the same typography as one resting on 590 million searches, and a modelled forecast reads exactly like a measurement.

Two such indexes prompted this page. One collects 145 figures from a single agency's own work; the other collects 138 findings from third-party publishers. Both are honestly built, both link each figure back to its source, and neither is a study. Neither has a denominator of its own, which is why neither is graded on this site's evidence scale — an index cannot be correct or incorrect in the way a study can. What can be assessed is the corpus they point at.

One caution about counting. The first index describes itself as drawing on 21 studies, tests and case studies; its own machine-readable data resolves to sixteen source URLs, of which eight are client case studies and three are separate write-ups of a single dataset. The second describes 28 studies, but seven of the sources are the publisher's own posts, leaving 22 genuinely third-party. Neither discrepancy is deceptive. Both illustrate that a study count is itself a number with a definition behind it.

Nine companies published seventeen of the twenty-two studies

Of the 22 third-party studies in the industry index, 17 were published by companies that sell a product measuring the same quantity the study reports as alarming. Ahrefs, Semrush, Similarweb, Profound, Evertune, BrightEdge, BrightLocal, Local Falcon and SOCi all sell AI visibility tracking in some form. The five remaining come from SparkToro, iPullRank, Seer Interactive, Sterling Sky and the Pew Research Center, and of those only Pew has no commercial interest in the subject at all.

Grid of 22 squares representing the 22 third-party studies indexed in this field. 17 are shaded to show they were published by companies that sell a product measuring the same quantity the study reports. Only 5 were published by organisations with no such product to sell.Who publishes AI search research22 third-party studies, by whether the publisher sells the measurement.17sell a product that measures this5do notThe 22 third-party studies indexed by Steady Demand,September 2026. The 17 come from nine companies.
Publishers of the 22 third-party studies in the industry index, by whether they sell a product measuring the same quantity.

This is not an accusation and disclosed funding is not disqualifying. The vendors say who they are, and several of them publish more carefully than the people who quote them. But funding decides scope, and scope decides findings. A company selling prompt tracking measures prompt sets. A company selling backlink data measures links. A company selling local visibility scans measures the businesses already in its scanning database. In every case the question worth asking is what the study would have found had it been built to look somewhere else, and in most cases the answer is unknowable from outside.

One panel, three sources

The independence problem is sharper than the funding problem, because it is invisible in a list.

SparkToro's zero-click study draws its data from Similarweb's clickstream panel. Similarweb separately publishes two of its own studies in the same index. Presented as three rows in a bibliography, these look like three sources converging. They are one panel and three interpretations of it, and the interpretations do not agree: the same underlying clickstream supports a figure near 37% in Similarweb's own published work and 68.01% in SparkToro's, because the two draw the boundary of a click in different places.

Convergence between sources is the main reason anyone trusts a number. When the sources share a panel, that convergence is an artifact of the shared input, and an index is exactly the format in which the sharing disappears from view.

Where two vendors disagree about the same measurable fact

The clearest test of a field's evidential health is what happens when two parties measure one thing that is not in dispute.

Similarweb reports that 26% of ChatGPT responses contain an advertisement. Evertune reports roughly one in eight, which is about 12.5%. This is not a question of definition or intent; either an ad is present in a response or it is not, and it is directly countable. The two figures differ by a factor of two, neither publisher discloses the method that produced its number, and the industry quotes whichever one suits the argument being made. When a plainly countable fact cannot be pinned down within a factor of two, the more interpretive figures in the same corpus deserve considerably less confidence than they usually get.

A similar spread appears on Reddit's citation share, which runs from around 2% in one large census to 13.7% in a study of local commercial queries. That particular gap is explicable rather than contradictory, because the two measured different query universes. Which is itself the recurring lesson: no AI surface publishes a query universe, so every citation-share percentage in this field is a share of a prompt list somebody wrote.

Errors travel further than the studies they come from

Indexes propagate mistakes efficiently, because the reader who meets a figure in a list has no method in front of them to check it against.

A worked example sits in the industry index itself. It reports that 5.4% of the restaurants ChatGPT recommended are rated below 3.5 stars, and attributes this to Local Falcon's Restaurant AI Visibility Index. That study never measured ChatGPT. Its scope is Google AI Overviews and Google Maps, and the 5.4% figure describes Google's AI recommendations. The number is accurate and the engine is wrong, which is the most durable kind of error: nothing about the figure looks suspicious, so nothing prompts anyone to check it.

The same mechanism operates at larger scale on claims rather than errors. When one widely repeated finding about ChatGPT's local data sources was traced back, its origin was 50 prompts across five Spanish cities — and seven of the eight downstream write-ups had stripped every caveat within nine days of publication. The caveat is always the first thing lost, because the caveat is the part that makes the claim less useful.

How to read a research index without being misled

Four questions handle most of it, and all four are answerable in a couple of minutes from the primary source.

What is the denominator? Almost every misleading figure in this field is a correct percentage of something other than what the reader assumes. A share of citations is not a share of queries. A share of tracked businesses is not a share of businesses. A share of a keyword database is not a share of searches.

Is the comparison observed or modelled? A decline measured against a forecast is a projection dressed as a measurement, and the difference is invisible once the number is in a list.

Does the publisher sell the thing being measured? Not to dismiss the work, but to predict what it did not measure.

How many prompts? Ask before accepting any claim about what an AI engine does in general. The answer is often two orders of magnitude smaller than the confidence of the claim implies.

Where a study on this site has been read in full against its own design, it gets its own page under the studies section. This page covers the rest of the corpus, and its purpose is narrower: to record what each number is a share of, so that a figure met in a list can be returned to the design that produced it.

What each widely quoted number is a share of
Published byThe headlineWhat the number actually counts
SparkToroIn 2026, less than one third of Google searches send a click68.01% of Google searches end without a clickA share of searches in a Similarweb panel, January to April 2026, in which a click to Google Maps or YouTube is counted as a zero-click. The same measure sat near 49% in 2019, five years before AI Overviews existed.
iPullRankWhat 13 billion Google searches reveal about zero-click behavior46.96% of search sessions end without a clickA session-level share of a clickstream panel that is 94.5% desktop, in a study whose own figures put mobile zero-click at 66% against desktop at 49%. The panel skew moves the headline.
AhrefsAI Overviews reduce clicks by 58%AI Overviews cut clicks by 58%Observed position-one CTR of 0.016 against a forecast of 0.037, built by applying a non-AIO cohort's two-year decline to the AIO cohort. It is a modelled counterfactual, not an observed before and after, and nothing in the design separates AI Overviews from everything else that changed between December 2023 and December 2025.
AhrefsInsights from 55.8M AI Overviews across 590M searches9.46% of keywords return an AI OverviewA share of a desktop, logged-out keyword index that excludes low-volume queries. The publisher states plainly that it undercounts. The 16% US figure from the same study is the one usually substituted when a larger number is wanted.
AhrefsTop brand visibility factors in ChatGPT, AI Mode and AI OverviewsYouTube mentions correlate 0.737 with AI visibilityA Spearman correlation across 75,000 brands pre-selected for domain rating above 40 and a high-volume head keyword — a population in which YouTube presence and AI mention both co-vary with simply being large. The prompt set is not disclosed, so it cannot be reproduced.
SemrushThe ghost citations study62% of AI citations never mention the brandA share of 3,981 domain appearances generated from 115 prompts. No date range is given, and the method used to detect a citation versus a mention is neither described nor validated.
SemrushChatGPT traffic analysis: 17 months of clickstream dataChatGPT referrals grew 206% year on yearGrowth measured on an undisclosed base within Semrush's own clickstream panel. The more useful figure in the same study is rarely quoted: 21.6% of ChatGPT's outbound referrals go to Google.
Seer InteractiveAIO impact on Google CTR: 2026 updateBeing cited in an AI Overview lifts CTR by 120%The difference between cited and uncited queries across 53 brands. The study states outright that it cannot support the causal reading, and notes that higher-authority brands are likelier to be cited in the first place. AI Overview status is also applied backwards from a recent SERP snapshot across fourteen months of data.
Local FalconThe AI visibility crisis83% of restaurants are invisible on ChatGPTA share of results for businesses already paying to be tracked in Local Falcon's own platform, which is not a random draw from US restaurants. The restaurant count and the prompt wording are not published.
Local FalconThe restaurant AI visibility index74.9% of restaurants are invisible in AI search“AI search” here means Google AI Overviews only — no ChatGPT, no AI Mode, no Perplexity. “Invisible” means absent from all nine points of a grid drawn around the restaurant's own address. The 19.6% Maps comparison sets a conversational prompt against a “near me” query, which are different surfaces and different intents.
ProfoundThe parrot problem47% of AI response content is unsolicitedDrawn from 50,000 prompts, with no stated denominator for the 47% itself, no date range, no model versions, no industry list and no description of the classifier that decided what counted as unsolicited.
SimilarwebAI search stats 202626% of ChatGPT responses contain an adNo method is published on the page; the methodology sits behind a gated report. Evertune puts the same measurable fact at roughly one in eight. Two vendors, a two-fold disagreement, neither showing its work.
BrightEdgeGemini becomes the No. 2 consumer AI referral sourceGemini reached 13.2% of AI referralsFigures reported to one decimal place with no disclosed sample, no site count, no traffic volume and no data source. It is a press release for a product, and it is cited in trade press as market data.
BrightLocalHalf of consumers are asking AI for business recommendations45% use AI for local recommendations, up from 6%A 455-person subsample of a 1,002-person survey. A 7.5-fold move in one year is more readily explained by a change in question wording or placement than by consumer behaviour, and the wording for neither year is published. No margin of error is stated.
Pew Research CenterAmericans and AI 202649% of US adults have used an AI chatbotA probability-based panel of 5,119 US adults, fielded 17 to 23 February 2026 and weighted to the adult population, with the field dates and method published. This is the control case: the only source here with no product to sell and a disclosed sampling design.
Questions

Reading the research

Is a research index a reliable source?

It is a reliable bibliography and not a body of evidence. An index collects figures away from the designs that produced them, so a number resting on 115 prompts is typeset identically to one resting on hundreds of millions of searches. Use an index to find primary sources, then read the method before quoting a figure.

How much AI search research is published by companies selling AI visibility tools?

Of the 22 third-party studies in the industry index reviewed here, 17 were published by companies selling a product that measures the same quantity the study reports. That is nine companies. Disclosed funding is not disqualifying, but it decides what gets measured and what is left unmeasured.

Why do zero-click studies disagree so much?

Mostly because of definitions and panel composition. One counts a click to Google Maps or YouTube as a zero-click and reports 68.01%; another measures at session level on a 94.5% desktop panel and reports 46.96%. Both are internally correct. They are not measuring the same thing, and neither supports the claim that AI caused the difference.

What does it mean that a study is a modelled counterfactual?

It means the comparison point was calculated rather than observed. The widely quoted finding that AI Overviews reduce clicks by 58% compares an observed click-through rate against a forecast of what it would have been, built by applying another cohort's decline. It is a projection, and it is quoted as a measurement.

Which AI search research has no commercial interest behind it?

Very little. In the corpus reviewed here the clearest case is the Pew Research Center's work on AI adoption, which uses a probability-based panel of 5,119 US adults with published field dates and weighting. It is the useful benchmark for judging what disclosure should look like.

Should a figure from a vendor study be ignored?

No. Several vendor studies are more carefully written than the coverage that follows them, and some disclose limitations their citers remove. The practical rule is to quote the finding at the precision its method supports, name the denominator when quoting it, and never treat a single vendor's number as settled when no second party has measured the same thing the same way.

Top