Aggregate at the edge
Count day, public path, actor class, bot family, source, verification label, and last-seen time. Never retain IP addresses, full user agents, cookies, or full referrer URLs.
PUBLIC AI DISCOVERY MEASUREMENT
Evidence
before hype.
A live, privacy-preserving view of which public evidence gets requested, which AI platforms send people, and whether fixed questions produce observable citations over time.
30-day request window40 fixed citation promptsNo invented baseline
01
LIVE SIGNALS
Three signals. Three different meanings.
Crawling measures discovery. AI user agents measure retrieval requested on someone’s behalf. Referrals measure people arriving from AI products. Only a verified answer check measures a citation.
Loading the public ledger.
02
REPRODUCIBLE METHOD
A result another person can challenge.
Use fixed prompts
Run the same 40 relevant questions across four AI products. Record every dated result, including failures and missing citations.
Respect the boundary
Never convert a bot hit into a citation claim. Never call an empty baseline zero. Never promise indexing, ranking, referral traffic, or recommendation. Crawler access does not guarantee any of those outcomes.
03
WHY THESE LEVERS
Current evidence, with limits attached.
OpenAI states that OAI-SearchBot controls inclusion in ChatGPT search results. Anthropic and Perplexity publish separate search and user-agent crawler identities.
OpenAI crawler documentationAhrefs found that YouTube mentions and branded web mentions had the strongest correlations with AI brand visibility in its study. Correlation is not proof of causation.
Ahrefs studyAhrefs reported AI referrals at 0.28% of total traffic across 74,752 sites in March 2026. The channel is real, but often small enough that careful instrumentation matters.
Ahrefs referral analysisGoogle says normal search fundamentals and valuable original content apply to AI features. It also says llms.txt has no effect on Google AI visibility.
Google AI search guidance