The Agenda We Use for the Monthly AI Visibility Review
A practical agenda for the monthly client review of an AI visibility program: what we show, what we skip, and how we leave with decisions.
Monthly review meetings have a way of turning into slideshows. Someone walks through twenty charts, the client nods, and nothing is decided. We have sat through enough of those on both sides of the table to design ours differently.
Our monthly AI visibility review is forty-five minutes. It has a fixed order, a short deck, and one rule: every section ends with a decision or a clear statement that no decision is needed.
Before the meeting
We send a one-page summary two days ahead. It has the headline movement, the three things we want decided, and a link to the full dashboard for anyone who wants to look. Attendees who read it arrive ready to talk. Those who do not still get the same content in the first five minutes.
We prepare the data from the same prompt panel as last month. We confirm that nothing in the definitions has changed, and if it has, we flag it on the first slide.
In Promptwatch we usually produce the report as a scheduled PDF of the relevant dashboard, so the data in the deck matches what the client can open themselves. Custom dashboards let us arrange it the way this client thinks.
The agenda
1. What happened (ten minutes). Visibility Score, Share of Voice, Sentiment, and Position in Answer over the period, against the previous one. We show the split by engine and by prompt type. We state the caveat on variance in one sentence. We do not narrate every number.
2. Why it happened (ten minutes). Citations, crawler data, and competitor movement. Which pages gained or lost citations? Were there access errors? Did a competitor publish something that took ground? This is where the platform's linkage of crawler logs and citations earns its keep: we can usually give a specific reason instead of a guess.
3. What the traffic did (five minutes). AI-referred visits and conversions from visitor analytics. Honest framing of size. A comparison with last month. Pages that received the visits.
4. What we did (five minutes). The work shipped since last time: pages published or refreshed, access fixes, offsite outreach. No padding. If we shipped less than planned, we say why.
5. What we plan next (ten minutes). The next month's priorities from the Action Board, with a rationale for each. We bring three to five items, not twenty. The client chooses what to approve, defer, or drop.
6. Decisions and owners (five minutes). We read back the decisions, name the owner of each, and agree dates. Then we send the notes the same day.
What we leave out
- Vanity charts. If a chart does not support a decision, it stays in the appendix.
- Daily fluctuations.
- Lists of every prompt. We show clusters.
- Competitor trivia with no action attached.
- Tool tours. Clients care about outcomes, not interface.
How we handle bad months
Sometimes the numbers drop. We lead with the drop, not bury it. Our approach:
- State the fact.
- Say what we checked and in what order: panel changes, competitor activity, client changes, access, engine behaviour.
- Say what we found, or that we have not found it yet.
- Propose a response.
The fifth item, engine behaviour, comes last because it is the easiest excuse. If we cannot find another cause, we say so and plan to watch.
How we handle good months
We are cautious about taking credit. A rise might follow our work or a change on the engine's side. We describe what we did and what moved, and say that the link is plausible rather than proven when the evidence is not strong. Over time, repeated patterns build confidence, and clients trust a measured voice more than a boastful one.
Who should attend
The person who owns the budget, the person who owns the website, and whoever approves content. If engineering matters, such as crawler access issues, a representative should attend at least occasionally. Meetings where the people who can make the decision are missing become relay races.
Using the Action Board as the spine
We keep the next-month list in Promptwatch's Action Board, which gives prioritized GEO tasks, assignees, and comments. The weekly Discovery run suggests evidence-backed actions, up to five per run. We bring a subset to the meeting. Clients appreciate seeing the reasoning attached to each item, and the board doubles as the record of what was decided.
Quarterly variation
Every third meeting we step back. We review the prompt panel, the competitor set, and the definitions. We ask whether the program's goals still match the business. We look at new engines or features that might deserve inclusion. And we revisit the scope, because scope creep is a cost we prefer to manage openly.
Follow-up
Within a day: notes with decisions and owners. Within a week: first deliverables from the approved items. The following month's meeting starts by checking what was promised against what happened. If you make a habit of closing the loop, the meetings get shorter.
Our recommendation
Keep the review short, put decisions at the end of every section, and make the data traceable. We build the pack in Promptwatch, using dashboards, scheduled PDF reports, and the Action Board, so what we show matches what the client can verify. You can see the reporting features at promptwatch.com.