Traffic charts answer the wrong question
Sessions are increasingly a poor proxy for commercial performance, since informational visits are disappearing while enquiries hold. A report leading with them invites the wrong reaction.
Section 1: appearance against a fixed prompt set
How often you appeared, across which engines, for the same prompts as last month. The prompt set must not change during a reporting period, or the trend is meaningless — the core measurement discipline.
Section 2: how you were described
Actual sentences the engines used about you, with any inaccuracies flagged. This is the section clients find most persuasive, because it is verifiable in thirty seconds and it converts abstract visibility into something concrete.
Show them the sentence a model wrote about their company. Nothing else in the report lands like that.
Section 3: what changed and what it did
The specific pages published or restructured, and what moved after. Honest attribution here — including “no measurable change yet” — is what keeps the report credible when the slow levers take quarters.
Section 4: the next actions, ranked
Questions to answer, facts to publish, sources to be present in. A report that generates no backlog is theatre, and this is where a competitor audit turns into work.
Monthly, not weekly
Answers vary between runs, so short intervals report noise. A monthly cadence with repeated sampling per prompt is the shortest period that produces a signal worth acting on.