Free tool - no login
Paste an answer from ChatGPT, Perplexity, Claude, or Gemini and your domain. We parse it for whether the engine named your brand, cited your URL, and which sources and other brands it leaned on. The full audit does this automatically across many prompts and engines, and tracks it over time.
A citation is a link to your URL inside an AI answer; a mention names your brand without linking. Both matter, and they diverge - an engine can know a brand well and still not link it. Paste an answer and your domain here to classify exactly what the engine did: named you, cited you, or leaned on other sources entirely.
Free and instant. You supply the answer text, so there is no engine call.
Four methods, in rising rigor. Ask the engines your real buyer prompts and inspect the sources they attach. Check server logs for OAI-SearchBot and ChatGPT-User visits. Watch analytics for AI referral rows, plus Bing Webmaster Tools' AI performance view - a useful window since ChatGPT search leans on Bing's index. And run the same prompts repeatedly on a schedule, because one run is an anecdote.
Both, for different reasons. Mentions shape the shortlist a buyer forms before visiting anyone; citations carry attribution and the click. When engines name you without linking you, you are one strong, extractable page away from converting mentions into citations - which makes that gap a precise optimization target.
This tool parses a single response. The free audit runs your whole prompt battery across five engines, stores every answer as evidence, and charts who wins each prompt.
Because engine answers are not deterministic: sampling temperature, model routing, and live retrieval all move the output between runs. Presence is a rate measured over repeated runs, not a screenshot - and trends require a fixed baseline to be readable at all.
Baselines and scheduled re-tests are exactly what the Proof Loop exists for.
Because the failure lives in one of three layers with three different fixes. Access: AI bots blocked at robots.txt or the CDN. Extraction: the answer exists but lives in client-side JavaScript or buried prose. Grounding: engines can read you but select sources they trust more.
Classifying the miss matters more than counting it - the fix for a blocked bot and the fix for a grounding loss have nothing in common.
And when citations do land, Traffic Lens surfaces the AI-driven sessions your analytics buries in Direct.
It reads one pasted answer. It cannot sample engines for you, and it cannot watch movement over time.
The free audit automates the sampling across five engines; the Proof Loop re-tests tracked prompts on a schedule and emails you when a citation lands.
A citation is a source link to your domain in the answer. A mention is your brand named without a link. Mentions build the shortlist; citations carry the click - and the gap between the two is itself a fixable target.
Look for the user agents OAI-SearchBot and ChatGPT-User for OpenAI, PerplexityBot for Perplexity, and Claude-User or Claude-SearchBot for Anthropic. Frequent hits on a page signal it is in rotation as a citation candidate.
Partially. GA4 referral rows for AI domains - and its AI-source groupings - count clicks only, so treat them as a floor: many answers inform a buyer without ever sending a visit.
No - sourcing differs sharply between engines, so a conclusion drawn from one engine says little about the others. Check several before deciding anything.
Find the prompts you lose, ship a targeted fix, and re-test the same prompt against a fixed baseline. That is the Find, Fix, Prove loop the platform automates end to end.
Explore the other free checks: