AI visibility reporting for agencies: what belongs in a monthly client report

Open with the answer to 'how are we doing?'

Start the report with a short summary in plain words: where the brand stands, the one or two biggest changes, and what you recommend. A client who reads only this section should still know what happened.

Support it with a small number of headline figures: visibility in unbranded prompts, share of voice against the main competitors, and the share of positive claims in branded prompts. Show each with the number of answers behind it.

Show branded and unbranded separately

Branded prompts are about reputation: what a buyer is told when they check the client by name. Unbranded prompts are about discovery: whether the client is named when nobody asked. Combining them hides both stories. See our guide to branded vs unbranded prompts for the reasoning.

For branded, list what engines praise and what they doubt, with real example sentences. For unbranded, group questions by type and show average visibility per group, so the client sees where they are strong and where they are absent.

Related reading: OClever vs sampled-prompt AEO tools for agencies and Why confidence intervals belong in every AI visibility report.

Make competitors concrete

Show who is named alongside or instead of the client, ranked by share of voice, for the main engines. A short table is enough. Pick out one competitor that gained and explain, from the answers, what the engines said about them and which sources they relied on.

Clients respond to specifics: 'Competitor A is recommended first in 6 of 10 comparison answers, because three review sites rank it highest on price' is a story. 'Competitor A has higher share of voice' is not.

Explain movement honestly

Generated answers vary, so a change of a few points on a small sample may be noise. State the sample size and, where you can, a range. Call out changes that are large enough, or sustained over several months, to be treated as real.

When something does move, say what you think caused it and how you know: a new page was cited, a competitor lost a source, a crawler block was lifted. If you do not know, say that, and say how you will find out.

End with three to five actions

Each action should be specific, tied to evidence in the report, and have an owner and a date. Examples: publish a pricing comparison page that answers a question competitors currently win; correct an outdated listing that engines repeat; unblock a crawler that is receiving errors; add a case study for the segment where visibility is zero.

Close the loop on last month's actions: what was done, and what changed as a result. Reports that review their own recommendations earn trust.

Operational tips for running many clients

Keep one prompt library structure for all clients so reports are comparable and quick to produce, then customise the prompts per client. Freeze each client's baseline set and add new prompts as separate cohorts.

Use consistent definitions across clients: what counts as a mention, which entities are counted as competitors, how branded prompts are tagged. Document them once and link to them in each report.

OClever is built for agencies managing several clients: a separate workspace for each client, a report page per brand with the branded and unbranded sections described above, and role-based access so clients see only their own data.