top of page

AI Search Visibility: What Three GEO Studies Show About Who Gets Named

Oct 23
7 min read

This analysis was produced by AI, TELL ME!, a full-cycle AEO & GEO agency with our own AI search monitoring platform.

Summary

A buyer who asks an AI assistant for a CRM (customer relationship management) tool, a heart medicine or a pair of skis meets the same few brand names, and AI search visibility in three GEO (generative engine optimisation) studies by AI, TELL ME! follows that pattern from the UK to Austria. Across 465 answers, the three most-named brands held between 37.7% and 63.5% of all brand mentions in their category, brand-owned websites supplied between 5.6% and 32.0% of the citations or citation weight behind those answers, and between 3 and 6 brands per category that sell in the market were named in no answer. The reading that fits all three studies is that the shortlist forms mostly on pages a brand does not own, and the brand's own site works as a threshold to clear: among named brands, no site readiness measure showed a detectable link to how often a brand was named. The sections below set out the evidence, state what the different set-ups allow and turn the findings into checks for any category.

Cover titled AI Search Visibility: three GEO studies, one pattern. The top three brands hold 61.3% of mentions for CRM software in the UK, 37.7% for pharma companies in the UK and 63.5% for ski brands in Austria, against equal-split baselines of 8.1%, 13.6% and 25.0%.

AI, TELL ME! has published three GEO studies since September 2026: ski brands in Austria, CRM software for UK small businesses and pharma companies for heart and immune conditions in the UK. Each study put buyer-style questions to five or six AI engines and counted which brands the answers named, which websites they cited and how the brands' own sites looked to an AI crawler. Side by side, they show which findings repeat across LLMs (large language models) and industries and which belong to a single category.

AI search visibility concentrates on a few brands in each category

In all three studies the three most-named brands held between 2.5 and 7.6 times the share of voice an equal split would give them. Pipedrive, HubSpot and Zoho CRM held 61.3% of the CRM mentions (summed from the study's table). Novartis, Eli Lilly and AstraZeneca held 37.7% of the pharma mentions, and Atomic, Salomon and Rossignol held 63.5% of the ski mentions. An equal split across the brands each study named would give the top three 8.1%, 13.6% and 25.0%.

Table 1. The three studies side by side (the CRM top-three share is derived from the study's table).

Measure

CRM, UK

Pharma, UK

Ski, Austria

Answers and engines

225, 5 engines

120, 6 engines

120, 6 engines

Brands named

37

22

12

Top three share of mentions

61.3%

37.7%

63.5%

Brands named in no answer

3 of 40

3 of 25

6 of 18

Brand-owned share of citations

32.0%

5.6%

10.3% (by weight)

The tail below the leaders is long in each category. In the CRM study 23 of the 37 named brands held under 1% of mentions each; in the pharma study 9 of the 22 named companies appeared in three answers or fewer; in the ski study 4 of the 12 named brands shared 2.3% of mentions.

Most of the sources behind AI answers sit on websites a brand does not control

Brand-owned websites supplied 32.0% of the citations in the CRM study and 10.3% of the citation weight in the ski study, and sites named after one of the 22 pharma companies supplied 5.6% of the cited items in the pharma study. Other parties supply the rest: in the CRM study, other companies' sites and independent sources carried about two thirds of citations; in the pharma study, government bodies, journals and guideline bodies supplied 66.8% of the 1,040 cited items; in the ski study, retail and marketplace domains carried more citation weight than every manufacturer site combined.

A short list of domains carries much of the sourcing in every category. The ten most-cited domains held 33.3% of the CRM citations and 42.5% of the ski citation weight, and the 15 most-cited domains held 50.5% of the pharma cited items. A mental health audit by Phuong Anh Nguyen and colleagues found the same shape in 15,942 citations from three consumer AI products: the top ten domains held 43.6% of English citations.

A reachable website did not guarantee a mention

Twelve brands that sell in their markets were named in no answer: 3 of 40 CRM brands, 3 of 25 pharma companies and 6 of 18 ski brands. The three never-named CRM vendors and the three never-named pharma companies all had reachable websites, and one never-named ski manufacturer scored 100 on crawler access and held no citation in the source set. Other never-named ski manufacturers did have access problems, among them noindex directives on 7 of 9 sampled pages at one site.

Among the named brands, no site property tracked mentions in the CRM or pharma study: none of the 33 CRM tests and none of the 28 pharma tests reached p < 0.05. In the ski study, structured-data readiness correlated with share of voice at +0.31 (p = 0.30) across 13 sites. With 15 sites at most per test, the studies can neither confirm nor rule out a link.

Differences in set-up limit the comparison to patterns

The three studies differ in market, language (English for CRM and pharma, German for ski), prompt count (45, 20 and 20), engine count (five or six) and citation basis, so the comparison reads patterns and adds up no percentages. The ski study counts citation weight (the monitoring platform's own percentage for each cited source); the other two count citations.

Each study is a single run with 10 to 15 prompts per segment, and the shares are directional. Dmitrij Żatuchin's audit of 4,500 repeated responses found that recovery percentages depend on extraction, question selection and the size of the reference collection, which is why each study states its counting basis.

What the three studies add up to

The three studies add up to one reading: the shortlist forms mostly on pages a brand does not own, and the brand's own site is a threshold to clear. The leaders hold a multiple of an equal share in every category, and the pages behind the answers are specialist test sites and retailers in ski, public bodies and journals in pharma, and other companies' sites plus review, press and community pages in CRM.

The own site decides whether a brand can enter that list at all. A site that blocks crawlers, carries noindex directives or shows its text only after JavaScript runs has no path into an answer, and the never-named ski manufacturers included such sites. Past that threshold, readiness stopped separating named from never-named brands: the CRM and pharma never-named brands had reachable sites, and none of the 61 tests among named brands reached p < 0.05.

Own-site citations do track mentions, at a Spearman correlation of 0.56 across the 37 named CRM brands and 0.71 across the 22 named pharma companies. The studies don't show which comes first, so a leader's cited site may follow from its visibility as easily as cause it. The studies are single observational runs, so this conclusion describes the pattern in the data and tests no intervention.

Checks a brand can run on its own category

Each check below answers one finding from the studies.

Table 2. Findings and the check each one calls for.

Finding in the studies

Check for a brand's own category

The three leaders held 37.7% to 63.5% of mentions

Measure share of voice and compare it with the leaders

ActiveCampaign was named in 21 of 75 CRM automation answers and in none of 75 general ones; AstraZeneca in 35 of 60 cardiology answers and 3 of 60 immunology answers

Run the prompts separately for each kind of buyer question

Roche held 15.8% of one engine's pharma mentions and 1.8% of another's

Compare the engines one by one

Brand-owned sites supplied 5.6% to 32.0% of the sources

List the domains that carry the citations in the category

A blocked or unreadable site has no path into an answer

Confirm that crawlers can reach and read the brand's own site

The work then runs in four steps:

  1. Clear crawler access (server and CDN rules, robots.txt, noindex directives), which the ski study expects to show in citation data within weeks.

  2. Make pages readable to an extractor, with opening paragraphs that answer their heading and structured data.

  3. Place the brand on the domains the answers cite, which the ski study expects to take a quarter or more because it depends on other people publishing.

  4. Re-measure the same prompts on a fixed cycle, because a change in share of voice counts when the questions stay the same.

AI, TELL ME! works as a full-cycle AEO & GEO agency alongside a client's team, as a fully outsourced function or as a blend of the two. Third-Party GEO starts from the sources the answers cite and prioritises placements by relevance, authority, speed and cost.

FAQ

How concentrated is brand visibility in AI answers?

The three most-named brands held 61.3% of mentions in the CRM study, 37.7% in the pharma study and 63.5% in the ski study. An equal split across the brands each study named would give them 8.1%, 13.6% and 25.0%.

Do brands control the sources behind AI answers?

Brand-owned websites supplied a minority of the sources in all three studies, from 5.6% of cited items in the pharma study to 32.0% of citations in the CRM study.

Does a brand's own website matter for AI search visibility?

The own site works as a threshold: a site that blocks crawlers or carries noindex directives has no path into an answer. Past that point, none of the 61 site-property tests among named CRM and pharma brands reached p < 0.05, and most of the sources behind the answers sat on third-party pages.

Why is a brand with a reachable website named in no AI answer?

The studies identify no single cause. Several never-named ski manufacturers had crawler access or readability problems, while every never-named CRM and pharma brand had a reachable site.

Do AI engines name different brands for different questions and engines?

Yes. ActiveCampaign was named in 21 of 75 CRM answers on automation and in none of 75 general answers, and the combined share of Pipedrive and HubSpot ran from 39.3% of one engine's mentions to 56.8% of another's.

See which brands AI assistants name in your category with a free AEO Monitoring scan from AI, TELL ME!.

bottom of page