AI search monitoring
One audit is a snapshot. Monitoring is what makes it evidence.
AI answers move. A single measurement tells you today's state; only repeated measurement of the same frozen questions tells you whether anything you did actually changed them.
Founding pilot pricing while we validate. 14-day pilot, founder-run, cancel any time.
What has to stay constant for a trend to be real
- The question set is frozen and versioned. Change the questions and the trend is a property of your list, not of the market. Our question sets carry a version, and a change is a recorded event.
- The engines and models are recorded per run. A number that moved because a provider swapped the default model underneath you is not a result about your brand. Every observation stores the model requested and the model that answered.
- Failures never become absences. If an engine times out, that run is excluded from the denominator. A provider error must never show up in your report as "the brand was not recommended".
- Raw answers are kept. Every provider response is stored before it is scored. When a detection rule improves, we re-score the history offline — the past is re-derived, not re-purchased.
Why we do not sell daily by default
Daily checking multiplies provider cost by seven and, for most categories, changes no decision. Frequency is a cost lever, not a quality signal. If a category genuinely moves daily we will say so and price it accordingly; we will not use frequency as a way to make a plan look bigger.
What a monitoring cycle produces
Per-question state
Absent with a competitor named, present but not recommended, or recommended — on each engine separately.
Movement
What changed since the last run on the same frozen questions, with the runs behind each change.
Citation evidence
Which sources the answers drew on, and whether any of them were yours.
Directional competitive view
Which brands are being named most across the set. Ordering, not decimals — the competitor set is the part that moves between runs.
Common questions
How often should AI visibility be measured?
Weekly is enough for most brands. Daily measurement multiplies cost without changing decisions, because the underlying answers do not move that fast for most categories.
What makes a monitoring result comparable over time?
A frozen question set. If the questions change between runs, the trend is an artefact of the question list rather than a change in how AI answers. We version and freeze the set, and record every change.
Can monitoring prove a change worked?
Only if the same frozen question set is measured before and after, on the same engines, with failures excluded from the denominators. That is what our before/after comparison does. It shows what moved; it does not prove your action caused it.
What we actually measure
Every result comes from the official OpenAI, Google Gemini
and Perplexity APIs with web search or grounding enabled. In our data those
surfaces are recorded as openai_api_web_search,
gemini_api_google_search and perplexity_api_sonar, and reports name
them explicitly.
That is not the same thing as a logged-in person using the ChatGPT app. These are official interfaces with web search on, but they are measurement surfaces rather than consumer sessions: no personalisation, no memory, no app-only features. Results can differ from what any individual user sees, and we do not claim the two are identical. We would rather put that on the pricing page than in the small print.
AI visibility · AI visibility tool · For agencies
Founding pilot pricing while we validate. 14-day pilot, founder-run, cancel any time.