Founder AI citations
How to measure AI mentions
Key answer
Measure AI citations with a dated prompt log: the same buyer questions, the same engines, pass/fail for whether your entity is named, and a screenshot or quote of the answer. Do not buy a vanity “AI share of voice” number with no methodology. Anthroly’s public benchmark uses published customer case studies, not a secret panel.
Founder, Anthroly
Sarath Kavuru is the founder of Anthroly. He runs a human-led SEO and AEO practice that turns founder expertise into AI-search citations, authority content, and qualified conversations. Full bio
The scorecard we actually use
This is the spreadsheet columns from Anthroly diagnostics. It is intentionally boring. Boring is measurable.
| Field | Values | Why |
|---|---|---|
| Date | ISO date | Models change; undated tests are anecdotes |
| Engine | Overview / ChatGPT / Perplexity / Claude | Do not blend scores |
| Prompt | Exact string | Wording changes answers |
| Named? | Yes / No / Partial | Partial = category described, brand missing |
| URL cited? | Yes / No | Mention ≠ citation |
| Accuracy | Correct / Mixed / Wrong | Wrong descriptions are worse than silence |
| Next URL | Path to ship | Closes the loop |
What not to trust
Single-prompt screenshots on LinkedIn. Tools that output a percentage with no prompt list. Conflating branded Google queries with ChatGPT mentions. Our benchmark page states its sample: published case studies, not a census of the web.
Frequently asked questions
How often should we re-run prompts?
Monthly is enough for most founders. Weekly if you are in a fast news niche (security incidents) and you are shipping pages that often.
Is referral traffic from ChatGPT showing in analytics?
Sometimes, with ugly referrers or none. Treat analytics as a lagging signal. The prompt log is the leading one.