Buyers ask AI for recommendations now. Mentioned asks Claude 5 real buyer questions about your category and checks whether you come up — free, with the real quotes to prove it. Then track those same five questions week over week, across Claude, ChatGPT, Gemini, DeepSeek and Llama, on a dashboard where every number links to the answer it came from.
The assistants Mentioned asks
Claude runs on our budget, free. The other four run on an API key you paste — used for that one check and never stored, or saved encrypted so a tracked keyword can be re-asked on them. Gemini and Groq both have free tiers, so a second opinion need not cost anything.
Add a keyword and Mentioned re-asks the same five questions on a schedule, so a move in the line is a move in your visibility — not a different question being asked.
How often AI assistants name Pickleland when a buyer asks about the category — measured on frozen questions, so week over week means one thing.
Share of answers naming each brand, last 12 weeks. Days where nothing was measurable break the line rather than being drawn through.
Sample data for a demo account — nothing here is a real customer’s result. The screen is real: it is the dashboard, rebuilt in HTML from the same stylesheet the product uses. See all nine screens →
Each card below is one claim and the piece of the interface that proves it — the real components, at the size they actually render.
The same five questions every run, so a rise is a rise in visibility — not a different question being asked.
Being named tenth is not being named first. “Not named” is counted here rather than dropped — it is the denominator.
Your score is five questions averaged. This is the half an average hides — ranked weakest first.
Beginner league night?
|
0% |
Indoor when it’s raining?
|
27% |
Best indoor courts in Austin?
|
93% |
Identical questions, asked of each assistant. A model at 0% is shown at 0% and never averaged with one that cites you.
| Claude | ChatGPT | |
|---|---|---|
| You | 84% | 41% |
| A. Pickle Ranch | 62% | 78% |
| Chicken N Pickle | 44% | 66% |
A cue count, not a model’s opinion — and the whole word list is on the screen, so you can check any verdict.
Answers are stored and printed verbatim, marked by the same matcher that scored the run — so a highlight can never contradict a score.
“Best indoor pickleball courts in Austin?” — Claude, Aug 24
Points lost times the answers it spans. Every row restates a measurement from another screen and links to it.
You can hold your score while the model starts naming twelve brands instead of three. This is the metric that catches it.
Never a list you typed. Every brand here is one an answer actually named, counted from the latest run of each keyword.
A measurement tool is mostly small choices about what to do when a number is missing. These are the ones that shaped this one.
Most tools rewrite their prompts each run, so part of every movement in their charts is the question changing. Yours are written once and reused verbatim.
“Nobody named you” and “nothing has run” are different findings, and this product never prints one as the other.
Filters and the trend window live in the URL, and the dashboard ships no JavaScript. Every view exports to CSV carrying the same filters.
IP addresses are stored hashed, never raw. A key you paste for a one-off check is used for that request and thrown away.
No scraping, no guessing — Mentioned reads the same answers a buyer would get.
Five realistic questions a buyer would ask about your category — not generic prompts, and never your brand name.
All five go to the assistant independently and simultaneously, exactly as a buyer would experience it — one question, one honest answer.
Each answer is checked for your brand, quoted where you appear, and read for who got recommended instead. No fabricated results, ever.
Track the keyword and those same five questions re-run weekly — or daily on your own key — building a trend that only moves when your visibility does.
Ranking well in Google does not guarantee a mention from Claude, and vice versa. These measure two different things.
A checker only matters if you can trust the result. Mentioned is built around that constraint, not around it.
Every mention, snippet, and competitor name is a real substring of the model's actual answer — nothing is hand-tuned, estimated, or invented.
Paying for higher limits buys quota, never a better score. The check runs the same for every brand, funded or not.
The free check runs on Claude, on us. ChatGPT, Gemini, DeepSeek and Llama are real, bring-your-own-key options — never claimed as live coverage we don't actually have.
Yes. Mentioned asks Claude 5 real buyer questions about your category and reports how often your brand comes up — with the actual quote when it does, and who got named instead when it doesn’t. It’s free with a free account, takes about 30 seconds, and gives you 5 checks a day.
An AI citation checker tests whether AI assistants like Claude and ChatGPT mention a specific brand when people ask for recommendations in that brand’s category. Mentioned asks 5 realistic buyer questions and reports how often your brand appears in the answers.
More buyers now ask AI assistants for recommendations instead of searching Google directly. If an AI doesn’t mention your brand for your category, it’s recommending competitors instead — even if you rank well in traditional search.
Yes. Mentioned is free, with a limit of 5 checks per day per account and up to 25 tracked keywords that re-run weekly. A paid plan for higher volume and more providers is on the pricing page.
The free check runs on Claude, paid for by us. You can also compare ChatGPT, Gemini, DeepSeek and Llama 3.3 by supplying your own API key for them — pasted into a one-off check and thrown away, or saved encrypted so a tracked keyword can be re-asked on that model. Gemini and Groq both have free API tiers, so a second provider need not cost anything.
Mentioned writes 5 realistic buyer questions for your category, asks the AI model each one independently, and checks whether your brand name appears in each answer. The score is simply how many of the 5 answers mentioned you. See how it works for the full methodology.
Yes. A tracked keyword is a brand and a category that re-runs on a schedule — weekly on our key, or daily if you have saved your own API key for every model it asks. The five questions are frozen when you create the keyword and reused verbatim on every run, so a move in the trend is a move in your visibility rather than a different question being asked. A free account can track up to 25 keywords. See the keywords screen.
Visibility is the share of AI answers that named your brand. Share of voice is your mentions as a share of every brand named in those same answers, which catches the case where you hold your score but the model starts naming twelve rivals instead of three. Average position is where you land when you are named — it is a rank, so lower is better, and it is left blank rather than shown as zero when you were named only in prose.
Yes. Save an API key for ChatGPT, Gemini, DeepSeek or Llama 3.3 and a tracked keyword’s frozen questions can be asked there too, then read back as a brands-by-models matrix. Because the questions are identical, a gap between two models is a gap between the models rather than a gap in what they were asked. A model that never names you is shown at 0% and never averaged with one that does. See the models screen.
It counts listed praise and criticism cues in the sentence that actually names your brand, and shows you every cue that fired and the sentence it fired on. It is not a model’s opinion of your brand and it is not a score out of ten. It will miss sarcasm and praise phrased in words that are not on the list — which is why the whole word list is on the screen rather than behind it. See the sentiment screen.
There’s no official “good” threshold — Mentioned doesn’t grade you against a benchmark. 5/5 means your brand showed up in every one of the 5 answers; 0/5 means it showed up in none. What matters more than the raw number is who got named instead when you didn’t — those are the competitors currently winning that answer — and how the score moves over time or across categories you check.
A rank tracker measures where a URL sits in a search engine’s results list for a query — a stable, repeatable position. Mentioned measures something different: whether an AI assistant actually names your brand in a generated answer to a real buyer question. AI answers aren’t a ranked list and can vary between runs, so ranking well in Google doesn’t guarantee a mention from Claude, and vice versa. See the full breakdown.
We store the check itself (brand, category, questions, answers and score) so a shareable results link works, plus a hashed version of your IP address for rate limiting. We never store raw IP addresses. See the privacy policy for details.
Run a free check now — then track it. A free account gives you 5 checks a day and up to 25 keywords re-asked weekly on frozen questions.