How to build an AI visibility dashboard on top of an API
Build a project per client, store prompt sets as reusable templates, schedule runs at a cadence matched to each client's budget, persist the full distribution rather than a point estimate, and define alerts that account for sampling uncertainty and repeated comparisons. Keep raw answers so any number can be evidenced on demand.
Nostimates supplies collected answers and search results through an API. Agencies and platforms can build their own dashboards on those records, with their own brand, data model and client access controls. The design below describes responsibilities for your application, not hosted Nostimates features.
Data model
- project → client: implement and test tenant isolation in your application. Keep the Nostimates account token on your server; it is not a client-scoped project key.
- prompt_set → a reusable template. Onboarding client #41 should take minutes, not an afternoon.
- run → prompt × surface × country × timestamp, storing the full distribution, not a single number.
- sample → the raw answer text, retained for evidence.
Cadence is a cost dial
Frequency is a 7x multiplier between weekly and daily. So is surface count, and so is n. Give your account managers a per-client dial and let them spend more on the clients who pay for it.
Alert off the interval, not the point
Firing an alert every time a rate moves produces noise your team learns to ignore. Define an alerting rule that accounts for sampling uncertainty, repeated comparisons and the size of the change that matters. An estimate outside an earlier interval alone is not a guarantee of a meaningful change.
Agree usage terms
Build your own analysis
Collect AI answers and search results for your own applications. Explore the public docs or request API access to discuss targets and volume.