When a buyer asks an AI assistant "what's the best tool for X", the answer names three or four products. Either you're one of them or you aren't, and nothing in Search Console or GA4 will tell you which.

Cogny now checks for you. Give it a question your buyers ask, your brand and your competitors, and it asks ChatGPT, Claude, Gemini and Perplexity, each with its own web search switched on, and scores what comes back.

What a check tells you

For each engine:

Mentioned or not, and how many times. Your brand matches on its name, any aliases you give it, your domain, and the domain's name without the TLD.
Rank: the order in which you and your competitors are first mentioned in the answer. "Mentioned third, after HubSpot and Pipedrive" is a different problem from "not mentioned".
Share of voice: your mentions as a share of all brand mentions in the answer.
Cited sources: every source the engine cited, and whether it's yours, a competitor's or a third party's.

Across the four engines you get a summary line: which engines mention you, your average rank, which engines cite your site, which competitors show up, and the most-cited domains.

The prompt goes in exactly as you write it, with no instructions from us around it. That's the point: you see what an ordinary user would see. Write prompts the way a buyer asks, without your brand name in them. "Best CRM for a 10-person agency" measures something. "Is Acme a good CRM" doesn't.

Tracking it over time

AI answers change from run to run, so one check is a snapshot, not a verdict. Every check is stored in your workspace, and a second tool reads the history back:

mention rate, cited rate, average rank and share of voice, overall, per engine, per prompt and per week
competitor mention rates across the same prompts
the domains the engines cite most, which are usually review sites, directories, communities and competitors' comparison pages

You don't need a data warehouse for any of this. The history lives in Cogny.

The weekly report

The AI Visibility Tracker report template runs this on a schedule. On its first run it builds a prompt set from what Cogny knows about your company: 8–10 buyer questions across category, problem, comparison and use-case prompts, plus your competitors. It saves the set in your context tree at ai-visibility/prompts. Edit that node to change prompts or competitors; the report keeps them fixed between runs so the weeks stay comparable.

Each week it reports:

Mention rate this week against last week, and the biggest change
A scoreboard per engine
Every prompt, the engines that mention you, and the competitor that leads the answer
Who wins the answers, and which sources the engines trust
Three to five actions, each tied to a prompt and a cited source. For example: get listed on the review site that every engine cites for your category, or write the comparison page a competitor already ranks with.

From your own agent

Both tools, check_ai_visibility and get_ai_visibility_history, are on your workspace's MCP endpoint. Any agent connected to Cogny can run a check in the middle of a task or pull the trend: Claude Code, Cursor, Claude Desktop, or your own scripts. Use it before and after you publish a page, or across a whole keyword list.

What it costs

AI Visibility is built in on every plan, with nothing to connect. Each check is billed in credits for the engines it asks, and an engine that fails to answer costs nothing. Ask fewer engines to spend less. Reading history is free.

Agent Mindshare is now part of Cogny

If you came here from agentmindshare.com: Agent Mindshare has merged into Cogny. Its free AI visibility preview stays on agentmindshare.com. Ongoing tracking across ChatGPT, Claude, Gemini and Perplexity is here now, together with the rest of your marketing data, so the fix for a missing mention can come from your Search Console, your content and your ads in the same place.

Start on Cogny →