Mentioned asks AI assistants the questions your buyers ask, the same way every week. You see where you’re recommended, who gets named instead and what the assistant says about them, and a ranked list of what to fix. This tour uses a made-up company, Northwind CRM, so every figure below is sample data.
The overview answers the question you were asked in the last planning meeting: are we on the shortlist, and who is on it instead? Then it opens on what to fix first, so the number comes with a next step.
12 keywords · 60 questions tracked, asked the same way every run
Claude mentions you in 9 of 12 answers to the same questions. See how to fix it
Mostly on “alternatives to Salesforce” questions, across all three assistants. See where
You’re mentioned first in 6 of the 9 answers that include you. See the answers
You were mentioned in 38% of answers in the latest run, up 4 points. You’re 3rd of 6 brands on share of voice, 17 points behind Salesforce.
| Brand | Latest | Change |
|---|---|---|
| Salesforce | 71% | −1 |
| HubSpot | 66% | +7 |
| Northwind CRM | 38% | +26 |
| Pipedrive | 26% | −1 |
Sample data for a made-up company, never a customer’s result. Salesforce, HubSpot, Pipedrive, Zoho CRM and Close appear as the kind of competitors a CRM would face; their figures are invented too.
Actions reads every run and lists the gaps worth your team’s time: a question you never win, an assistant that never names you, a competitor pulling ahead, a competitor quoted where you aren’t. Each finding links to the answers behind it. Under each one sits a short playbook of the usual fixes, labelled as advice so it is never confused with a measurement.
What’s costing Northwind CRM recommendations, ranked by the answers at stake. Every row shows the arithmetic behind its rank.
Salesforce appeared in 148 answers across 12 of your tracked keywords; Northwind CRM in 96 across 12. When both are named it averages #1.6 against your #3.4. 1 other brand is also ahead.
52 answers at stake · 35 points across 148 answersOpen the competitive map → + What usually moves thisHubSpot was named in 21 of 30 stored answers to “Which CRM is easiest to roll out without a consultant?” that never named Northwind CRM. This is what Claude said about it, word for word: “HubSpot is the easiest to roll out without outside help, with setup wizards and free onboarding for small teams.”
21 answers at stake · 70 points across 30 answersRead the full answer →The model has a specific, repeatable claim about your rival and no equivalent one about you. It isn’t ranking you lower. It has nothing to say.
“Which CRM is easiest to roll out without a consultant?” has named Northwind CRM in 0 of 30 answers. On the same keyword, “Which CRM has the best pipeline reporting?” named it in 19 of 30 (63%).
19 answers at stake · 63 points across 30 answersSee this keyword’s prompts → + What usually moves thisThe playbook is the one place in Mentioned that gives advice. It stays collapsed until you open it, and it never appears in a number, the CSV, the API or the MCP tools.
One card per screen: the question it answers for your team, and the panel that answers it. Every figure is sample data for Northwind CRM.
A keyword is your brand plus a category, like “CRM for B2B sales teams”. Each one runs weekly on our key. Pro adds 3 daily slots on our key and Agency 10, and any keyword can run daily on your own key. A free account tracks up to 5.
CRM for B2B sales teams Coreweekly | 35% | |
pipeline reporting software Reportingdaily | 52% | |
Salesforce alternatives for mid-market Alternativesweekly | 27% |
The five questions are locked when you create the keyword and asked the same way on every run. They never change, so when your number moves, it’s real. This screen lists them weakest first.
Every brand an answer actually named, never a watchlist you typed. Further right is named more often, higher is ranked better, and a bigger bubble appears in more of your keywords.
The same questions, broken out per assistant. A gap between two models is a gap between the models, never averaged away.
Positive, neutral or negative, from the words that decided it. Each column shows its count, so three sentences can’t pass for thirty. On Starter and above.
Every answer is stored word for word, with you and every rival highlighted. When an assistant recommends a competitor, you can read the claim it repeats about them, like “best for dashboards out of the box”, and see where it has nothing to say about you. That’s your content brief.
“Which CRM has the best pipeline reporting?” · Claude · Aug 24 · Named, rank 3
“Which CRM do startups switch to from spreadsheets?” · Gemini · Aug 24 · Not named
Only links a model returned with its answer, which today means Perplexity on your key. Assistants that answer from memory show none, and the screen says so.
Cited in 28 Perplexity answers
Connect Search Console and GA4, read-only, and see your search queries and the sessions that came from AI assistants next to your AI visibility. Fetched from Google when you open the page, never stored.
Mentioned runs an MCP server, so you can ask about your numbers in plain language from Claude and get an answer that matches the dashboard. Read-only, on Starter and above.
“Which locked question are we losing worst, and what do the answers say?”
Claude, ChatGPT and Gemini run on us. Save your own key for the other three to compare them, run a keyword daily on your budget, and, with Perplexity, fill in Sources.
Encrypted. We never show it again.
Three metrics, because each one catches a loss the other two miss. The definitions sit on the same page as the numbers.
Open any keyword to see visibility and rank on one chart, a row for each locked question, and the full answer behind every count. The questions never change, so when the line moves after your team ships a fix, you can show your CMO it’s real.
CRM for B2B sales teams · 5 questions locked 8 Jun · asked weekly on Claude and Gemini · 12 runs recorded
The right axis is inverted, rank 1 at the top, so up is better on both lines. Position is dashed because the two series are in different units.
The stored answer the top row was counted from, word for word, marked by the same matcher that scored it.
“Which CRM has the best pipeline reporting?” · Claude, Aug 24
An unlabelled reversed axis is a trap rather than a chart, so this one says so in words.
Connect Mentioned to Claude, ChatGPT or any MCP client and ask in plain language. Clients that draw cards show the answer as one: the same figures as the dashboard, with a link back to it. Press Fix with Claude and Claude pulls the evidence and drafts the fix, then asks before it re-runs anything.
An AI visibility number is only useful if nobody can poke a hole in it. These are the choices that make Mentioned’s hold up.
The screen, the brand and topic filters and the trend window all live in the URL. Bookmark the view you report from, or paste it into the deck review and everyone sees the same thing.
If a figure couldn’t be measured, you see a dash and a sentence saying why. You never present a 0% that was really a run that didn’t happen.
The file carries the same filters as the screen, so it drops straight into your own reporting.
If an assistant times out, the keyword retries at the next run instead of waiting seven days, and the failure is shown rather than hidden.
Paying buys more queries, assistants and runs. It never buys a better number or hides a competitor. Saved keys are encrypted, and raw IP addresses are never stored.
You’ll find these eventually, so here they are up front.
Mentioned shows where you’re losing recommendations and the usual fixes for each gap, labelled as advice. It doesn’t edit your site or your listings, and it doesn’t promise a placement or a ranking. What it does promise is that the next run asks the same questions, so you’ll see whether the change worked.
They are generated, not looked up in a ranked list. That is why a tracked keyword asks five questions, three times each, on every model, and why the trend is the number to read rather than any single run. It is also why the questions are locked: it removes the one source of variation we control.
The Sources screen lists only links a model returned with its answer. Today that is Perplexity, on a key you have saved. Claude, Gemini and the others answer from what they already know, so they report no sources, and the screen says so rather than guessing where an answer came from.
It counts listed praise and criticism words in the sentence that names you, and shows you every word that counted. It will miss sarcasm and praise phrased in words that aren’t on the list. Showing the whole list is the trade: you can see exactly why a sentence was called what it was called.
The only mail Mentioned sends is account mail you trigger: address verification, password resets and team invites. Nothing here promises a report or an alert. When that changes, this page changes with it.
3 checks on our key for all time, and 5 tracked queries run weekly on us. Every run spends real API budget, so a query we can’t afford this month is deferred rather than marked failed. Add your own API key and the free plan keeps checking indefinitely; a paid plan buys more on our key instead.
Start with one check on your own company. A free account gives you 3 checks on us and 5 tracked queries, run weekly on us. No card. Every check spends real API budget, so it needs an account.
No account yet? The free tools check your robots.txt, llms.txt and structured data without one. Weighing options? See how Mentioned compares.