AI visibility monitoring
A visibility score is only as good as the prompts behind it. We build your prompt set from what buyers actually ask, track it every day in the engines they use, and read the numbers with you until they turn into a list of work.
How we set it up and read it
The tool is the easy part. The prompt set and the reading are where most tracking goes wrong.
Prompt set
We track the questions that carry revenue, not the ones with search volume
There is no reliable search volume for prompts, so we build your prompt set from what buyers actually say: sales calls, reviews, support tickets and Search Console. We expand it into the fan-out queries an engine runs on its own and track it every day in the engines your buyers use, so the numbers describe your market rather than a keyword tool's guess at it.
Built from what your buyers already say
- Sales calls
- Customer interviews
- Reviews
- Support tickets
- Search Console
Tracked daily in the engines your buyers use
The numbers
Visibility, share of voice and sentiment, prompt by prompt
Every prompt gets its own line, so you can see which questions you already win and which ones you lose to someone else. We read visibility (how often you are named), citations and share of voice (how often your pages are the source), position, and who gets named in your place.
A visibility score on its own tells half the story. We read it next to the leads it produces: an AI option in your "how did you hear about us" field, AI-referral sessions in GA4, and branded search in Search Console.
Heyflow's tracked prompts in Peec AI. Each row is a question a buyer asks, with how often Heyflow is named, its share of voice and the sentiment of the answer. The work
The reading ends in a list of work, not a report
How this fits into AI Search →Tracking is only worth it if someone acts on it. For each prompt you lose, we look at the sources the engines cite and decide where the work goes: a page on your site, a platform like YouTube or a review site, or a third-party list you should be on.
What the tracking pointed us to for Heyflow: YouTube was a top source in its category. After we rewrote the titles and descriptions of videos Heyflow already had, Google's AI Overviews started embedding them in the answer. What comes out of a reading and goes on your roadmap
- Prompts you lose, and who wins them instead
- The sources engines cite for them
- Pages to fix, pages to write, platforms to be on
Questions about AI visibility monitoring
Which AI engines do you track?
ChatGPT, Perplexity, Gemini, Claude, Microsoft Copilot, and Google's AI Overviews and AI Mode, but the mix follows where your buyers actually search. A company whose customers live in the Microsoft ecosystem is far more likely to ask Copilot than Perplexity, so Copilot gets weighted accordingly.
How is a prompt set different from a keyword list?
A keyword list comes from search volume. A prompt set comes from how your buyers describe their problem out loud, on calls, in reviews, in support tickets, and it reads like a question someone would type into ChatGPT, with their context in it. Our guide on selecting prompts shows how we build one.
Can we book monitoring on its own?
It usually runs as the first part of our AI Search work, inside the same Growth Partnership, because the reading only pays off when someone acts on it. For something narrower we also run fixed-scope projects, priced separately.
