This practical sets out what an AI-visibility report should contain and how to structure the queries behind it, so the numbers survive a client’s questions. It builds on the labelling convention from Lesson 4.3 and the outputs of the query script.
What this lesson covers
- Designing the query set
- The structure of the report
- The method box
- Confidence labels
- Evidence from real answers
- Recommendations tied to the audit
The query set comes first. The lesson shows how to build it from the audit rather than from a keyword list: the entities and topics the business wants to be known for, written the way people actually ask, grouped by intent, with branded and unbranded versions kept apart and a known query included for calibration. Each topic is queried on its own, and every one is sampled before its aggregate is used.
The report itself has seven sections: a one-paragraph summary, the scope and method, the scoped view of who is cited alongside you, topic-level results with a confidence label on each, real answers that illustrate the findings, limitations, and recommendations that point at specific entities and pages in the gap matrix. The lesson provides a template, explains what belongs in each section and what must never appear without a label — an undated number, a topic result from default settings, or an AI search volume presented as traffic.
The template mirrors the outputs of the query script, so the scoped table, the topic results with their sample check, the calibration note and the method box are filled from data rather than typed by hand. What remains is the part only a person can write: what the findings mean for this business and which fixes they justify.
The lesson also covers presentation: how to show a share of citations without implying it is a share of traffic, how to show change over time when the underlying dataset moves monthly, and how much of the method belongs in the main body versus an appendix. A good report, it argues, is one a sceptical reader can check line by line, and one that would reach the same conclusions if someone else ran it next month with the same settings.
This practical continues Module 4 of The AI-powered Entity Audit. It suits consultants and in-house teams who report on AI visibility and want a format that is honest about what the data can support.