You cannot look up which prompts your brand ranks for the way you check keyword rankings, because no AI provider publishes prompt logs. What you can do is define the prompt space yourself: build a panel of the questions buyers plausibly ask, run it across providers, and read the results as a matrix showing exactly where you appear, where you trail competitors, and where you are absent entirely. Defined honestly, that panel is a better map of buyer intent than most keyword lists ever were.
Build the prompt taxonomy
Buyer prompts cluster into four families, and your panel needs all of them because brands rank wildly differently across the four.
- Recommendation: "best email platform for a small nonprofit". The discovery moment.
- Comparison: "Mailchimp vs ConvertKit for creators". The shortlist moment.
- Alternatives: "alternatives to Mailchimp". The switching moment, and often the highest intent of all.
- Pricing and fit: "affordable email tool with good automation for two people". The qualification moment.
- Write 5 to 10 phrasings per family in your buyers' vocabulary, not your positioning language. The panel concept is unpacked further in what prompt-level brand tracking is.
Reading the prompt matrix
Lay results out with prompts as rows and providers as columns; each cell records whether you were mentioned, at what position, and with what framing. This matrix is the "dashboard for AI brand mentions" people go searching for, and the honest answer is that every real dashboard, home-built or from a platform, is this table with a refresh schedule. Populate it with several runs per prompt, since AI answers vary between runs and a single-run matrix is a snapshot of noise.
Spotting the cluster where you always lose
Losses in the matrix are rarely random; they arrive in clusters, and the cluster is the diagnosis. Blank across the alternatives row means nothing on the retrievable web frames you as an alternative to the category leader. Blank down the ChatGPT column while Perplexity is healthy suggests weak training-data presence with decent live citability. Blank on pricing prompts usually means your pricing is unpublished or rendered in JavaScript that AI crawlers cannot read. One cluster, fixed deliberately, moves more than scattered effort across the whole matrix, and running the same matrix over your competitors shows whose clusters you are losing to.
Track position, not just presence
Within every mention, record where you land among the brands named. AI answers are effectively ranked shortlists, and first-named carries very different weight from sixth-named in a list the buyer skims once. Position is also the early-warning metric: brands usually slide down answers before they fall out of them, so a mention rate that looks stable can hide a position trend that is quietly going wrong. A free scan captures mention, position, and sentiment per prompt across five providers, which is this entire page as a product.
See your prompt matrix without building it
The free scan runs organic buyer prompts across five AI engines and shows you, prompt by prompt, where your brand appears, where it ranks, and where competitors own the answer.
Frequently asked questions
Is there a dashboard for AI brand mentions?
Yes, AI visibility platforms provide one, but understand what it is: a prompt-by-provider matrix populated by sampling, refreshed on a schedule. No provider exposes prompt logs, so no dashboard shows what users actually typed; it shows how the engines respond to a defined panel of buyer prompts. Judge any dashboard by its panel quality and methodology.
How do I track my brand's position in AI recommendations?
For every sampled response that names your brand, record your rank among all brands mentioned, then track average position per prompt family over time. Position shifts usually precede mention losses, making this the best early-warning metric in AI visibility measurement.
How many prompts define my brand's prompt space?
For most brands, 20 to 50 prompts spanning recommendation, comparison, alternatives, and pricing-fit families cover the space that matters commercially. Larger catalogs or multi-product companies need panels per product line. Depth counts as much as breadth: several runs per prompt is what makes the matrix trustworthy.
Related answers