Guide: what a report should include
What an AI visibility report should include
A good report says what it measured, how, what it could not measure and what to do, with the evidence in view. These are the pieces it must include, and the signs of a report you should not trust.
The category you were measured in
Everything depends on how your business was described: if you were measured in the wrong category, the results do not apply. The report must say which category it used and where it came from, and ideally let you correct it before measuring.
The status of each engine
How many answers were requested and received per engine. If an engine failed and the report does not say so, the numbers are incomplete. Totals must match that table.
Mentions separated by engine type
Live-search engines on one side and memory engines on the other. A single number that mixes them hides what matters.
Verified competitors and cited sources
Who AI recommends instead of you, checked against your offer (not any brand that shows up), and which sources it cites to decide. External mention actions come from there.
Page-by-page changes, without duplicating what exists
What to keep, modify and create, checked against your sitemap: proposing to "create" a page you already have is a common mistake. One proposed title per page.
Signs you should not trust it
It claims something does not exist without having checked it; it writes your method, results or prices for you; it gives a perfect structured-data score without checking it matches the visible text; or it does not say what it could not measure.