What 51 AI answers reveal: visible but not cited

A pilot across 17 B2B SaaS brands, three buyer questions each, one logged-out run. The headline: most brands get named, few get quoted.

1. Mention rate is not the constraint — citation is

41 of 51 answers (80%) named the brand being asked about. That looks like good news, and it is the part everyone reports. The constraint shows up one line later: only 16 answers (31%) cited the brand's own website as the source. When an answer names a brand but cites a third-party comparison page, the brand is borrowing reach from a page it does not control.

2. The zero-citation group is not the low scorers

8 of the 17 brands were cited in none of their three answers. Several of them scored 70+ and were named in every answer — the ranking table a buyer sees is fine, the evidence line is not.

3. The high-citation brands publish pages that answer the question

The brands that kept their citation share — WorkOS 3/3, Tally 1/3 with a strong score, Customer.io and Airtable and Mixpanel — have documentation or comparison pages that read like answers to buyer questions. The pattern is not "more content"; it is "content phrased as the answer the engine is looking for".

4. What this means for a content team

• Pick the ten questions buyers actually ask; map each to the page that answers it best.
• Publish the answer as a page, not as a homepage claim: title the page with the question.
• Run the question three times, not once, and record which source wins each time.
• Compare against three named competitors on every question — a mention with no citation is a warning, not a win.

← The full index, brand by brand · Method