Getting named in 46% of AI answers with a 50-question bank
Result: 46% named, 88% cited (Sept 2026)
What was tried
A fixed bank of about 50 questions in three groups: brand, user intent, competitor comparison. The bank goes to every major model each month, and an agent records which answers name the brand and which cite its pages. The questions themselves came from reading how users phrase things on Reddit, then writing pages for those phrasings.
What it cost
API calls for the monthly run and the content agents. No ads, no agency.
What came back
Named in 46% of answers in September, up 15 points from August. Cited as a source in 88%. A growing share of signups now arrives from ChatGPT and Gemini.
What we would do differently
Treat “cited but not named” as its own backlog. Those pages are already trusted by the model; they need the brand to be said out loud, which is mostly an off-site endorsement problem.