Across eight real runs in September 2026, the median company was named in 23.5% of the answers to its own buyers' questions; five of eight were under a third. The two UAE local services scored 68% and 89%, the six global-category companies 12% to 34%. A perfect readiness score did not predict visibility, and the most-cited source was a listing, directory, review site or rival's page in seven of eight runs.
The eight runs
| Company | Market | Answers | Named | Own site cited | Visibility | Readiness | Top rival, share | Most-cited source |
|---|---|---|---|---|---|---|---|---|
| AI visibility tracker (software) | US | 34 | 4 (12%) | 4 | 10 | 100 | a social listening tool, 38% | hubspot.com (vendor) |
| AI email marketing tool | US | 36 | 9 (25%) | 7 | 21 | 100 | Brevo, 42% | g2.com (review site) |
| OTT release tracker (consumer site) | US | 31 | 4 (13%) | 2 | 10 | 65 | JustWatch, 45% | google.com (listing) |
| Cloud-kitchen platform | UAE | 19 | 13 (68%) | 7 | 57 | 73 | Talabat, 90% | linkedin.com (community) |
| Gym chain | UAE | 28 | 25 (89%) | 23 | 83 | 85 | Fitness First, 32% | a rival's profile page (competitor) |
| Mobile banking app | US | 32 | 11 (34%) | 10 | 27 | 94 | Robinhood, 16% | google.com (editorial) |
| Stainless steel fabricator | US | 46 | 10 (22%) | 17 | 14 | 90 | Ashland Conveyor, 7% | thomasnet.com (directory) |
| AI agents platform | US | 51 | 8 (16%) | 5 | 12 | 85 | Voiceflow, 14% | youtube.com (video) |
Answers: engine-and-question pairs that returned an AI answer. Named: answers with the company in the text. Own site cited: answers that listed a page of the company's as a source. Visibility and readiness are the report's two scores, out of 100. The top rival is the company named most often across the same answers; its share is the percentage of answers naming it.
What the numbers say
The median company was named in 23.5% of the answers to its own buyers' questions. Sorted, the eight shares are 12, 13, 16, 22, 25, 34, 68 and 89 percent. Five of the eight were under a third. That is the shape of the problem: for most companies, most of the time a buyer asks an AI who to use, the answer is a list they are not on.
Local beats global. The two UAE local services were named in 68% and 89% of answers; the six companies in global categories ranged from 12% to 34%. A local question has a short list of possible answers and the engines read local sources for it. A category question like 'best AI email marketing software' is answered from roundups that name the same ten brands everywhere.
Readiness did not predict visibility. The two sites with a perfect readiness score had visibility of 10 and 21. The gym chain scored 85 on readiness and 83 on visibility; the OTT tracker 65 and 10. Readable is the floor; being named is decided on pages the company does not own.
The rival is not always the market leader. For the mobile banking app, the company itself led the share (34% against Robinhood's 16%). For the AI visibility tracker, the top 'rival' at 38% was a social listening tool named because its name is an ordinary word; that run is why the report now demands proof before a common-word brand counts. For the cloud-kitchen platform, Talabat was named in 90% of answers: not a rival in the strict sense, but the name every answer reaches for.
The most-cited source was rarely the company's own page. In three of eight runs it was a listing, directory or community page (a Google listing, Thomasnet, LinkedIn); in one a review site (G2, cited in 10 of 36 answers); in one a video platform; in three a vendor's, a rival's or a publisher's page. The engines write the answer from those pages. The fix that moves the number is on those pages, and the report names them.
Google AI Overviews answered the least. Across the eight runs it returned an AI answer for 5 of the 13 to 15 questions on average; ChatGPT and Perplexity answered nearly every one. A score built only on AI Overviews would be built on a third of the questions, which is why the report prints the per-engine count beside every percentage.
What was wrong on the sites themselves
Product or Service schema was missing on five of the eight sites; FAQ schema on five; Organization schema on two; llms.txt on one. Every site allowed the AI crawlers and every one had a sitemap and HTTPS. The technical work left for most companies is an afternoon: the three schema blocks and a forty-line text file, all of which the generator writes from the site's own pages.
How to read your own
Run the free audit and put your row next to these. Under 25% named is the median; the question is not the number but which questions you lost and to whom, which is the report's second section. Then the sources table: whichever page the engines cite most for your questions and never names you is the first thing to change.
Questions about the runs
Are these real companies?
Yes: eight finished runs from September 2026, named by trade because the numbers are theirs to publish, not ours. Each was 13 to 15 buyer questions on ChatGPT with web search, Perplexity and Google AI Overviews, with Gemini in two of the eight.
Why are two of them so much higher?
Both are UAE local services. A buyer asking 'best gym in Dubai Marina' gets local names, and the local chain with fifty branches is one of them on nearly every engine. The six software and consumer companies are competing in categories the whole world's pages describe, and the pages the engines read name the biggest brands first.
Does a readiness score of 100 mean anything?
It means the engines can read the site. Two of the eight scored 100 on readiness and 10 and 21 on visibility. Being readable is the floor; being named is decided on other people's pages.
What would change these numbers?
In every run the report's first fix was one of three things: a review or directory profile that the engines were already citing for a rival, a page on the company's own domain that answers a lost question plainly, or a schema block the site lacked. None of the eight needed a rebuild.
How we get your company into AI answers
- 01
Read the answers
Your buyers' questions, put to ChatGPT, Perplexity, Gemini and AI Overviews. Who is named, in what order, which pages each engine cites.
- 02
Find the sources
The pages the engines cited, ranked by how many of your questions they decide. That list is the work.
- 03
Publish the gap
We write the page that answers the question you are missing from and ship it to your domain.
- 04
Earn the citation
Listings claimed, the sites the engines already trust pitched, and the same questions asked again at day 30 until the answer names you.
Put your row next to these
A free audit on the same engines, three minutes.