What Does AI Say
About Your Brand?
Two real questions - "What is your brand?" and "Best tools for your category?" - asked live to ChatGPT, Claude, and Gemini via their APIs, with a quote-validated verdict per engine: named or not named, and the exact span when named. No signup.
This is one sample per engine at one moment, measured via the providers' APIs without web browsing - not the consumer apps, whose answers vary by session, web results, and personalization. A mention counts only with a verbatim quote from the answer containing your brand or domain. Results are cached and shared for 24 hours: one paid run per domain per day, for everyone.
How it works
- Brand and category from your site - we read your homepage for the brand name (og:site_name, then title); the category comes from your input or a one-line inference from your homepage. If neither works, the category query honestly reports not measured.
- The same pipeline the paid product runs - identical engine calls and the identical quote-validation gate as Citevera's bundled weekly monitoring. This tool is not a lookalike; it is one run of the real thing.
- Verdicts you can audit - when an engine names you, we show the exact verbatim span that proves it. When it does not, you see the answer excerpt that named someone else instead.
- Guardrails instead of fakery - a global daily spend ceiling on our keys; at capacity the tool says so and serves cached results rather than inventing anything.
Example result
A real run of this tool against our own domain on 2026-08-04 - including the unflattering parts, because honest measurement is the product:
Quote: "I don't have specific information about 'Citevera' in my training data." Named - because the question contains the brand - but the engine admits it does not know us. That distinction is why the tool tells you to read the excerpts on this query.
Gemini described a citation-generator tool for students - a different product. Textbook entity confusion, live: the engine has a guess where a grounded entity should be.
None of the three engines named Citevera in the category answer. Honest baseline - this is the number the visibility work exists to move.
Common reasons AI answers do not name a brand
- The engine has no grounded entity for you - your name, domain, and description never co-occur clearly enough. Start with an entity home page and Organization schema.
- Category answers are built from consensus sources - listicles, review sites, forums - and you appear in none of them.
- Your content is not extractable: no direct answers, no clear who-this-is-for statements, so there is nothing citable to quote.
- AI crawlers cannot reach you - robots.txt or WAF rules block the crawlers that would have read you.
- A better-established name collides with yours and absorbs the mention (entity confusion).
- You are genuinely newer than the model's knowledge and not yet in its retrieval sources - visibility work compounds from here.
Frequently asked questions
How does this check whether AI mentions my brand?
It runs two real queries - "What is <your brand>?" and "Best tools for <your category>?" - against the ChatGPT, Claude, and Gemini APIs on our keys, then applies a quote-validated gate: a mention only counts if the answer contains a verbatim span that itself names your brand or domain. No quote, no mention - the same evidence gate our paid monitoring uses.
Why does my result say it was measured earlier today?
Results are cached for 24 hours and shared across all users - one paid engine run per domain per day, globally. That keeps a free tool honest about cost without ever inventing a result: you either get a real measured answer or a real cached one, with its timestamp.
ChatGPT told me something different when I asked it myself. Which is right?
Both can be. We measure via the providers' APIs without web browsing; consumer apps add web results, session context, and personalization, and answers vary between runs even with identical inputs. One sample is a data point, not a verdict - that is exactly why tracking uses repeated samples over weeks.
What does "not measured" mean?
That our call to that engine failed or was not configured - our side, not yours. It is deliberately distinct from "not named", which is a real measured result. We never blend the two.
The tool says the daily capacity is spent. Why not just run my check anyway?
The tool has a global daily spend ceiling on our API keys. When it is reached, new checks pause until tomorrow rather than degrade into something fake - cached results keep serving, and nothing gets invented. Paid monitoring runs on its own budget and is unaffected.
How is the paid monitoring different from this check?
Same engines, same quote-validated gate - but weekly scheduled runs with multiple samples per run (so "cited in 2 of 3 samples" instead of a single coin flip), sentiment, trend history, and your own tracked prompts. It is included in every paid Citevera plan from $49/mo, running on our keys with nothing to configure.
Not named anywhere? The fix path starts with what engines can read: check crawler access, verify your llms.txt, then run the full readiness check. Background: why competitors get cited and you don't and monitoring AI brand mentions without a paid tool.
Built by Paul, founder of Citevera · Published 2026-08-04 · Updated 2026-08-04 · Runs the same quote-validated pipeline as Citevera monitoring