GEO Playbook ยท Reporting

What belongs in a monthly AI search report

The short version

  • A report is a story, not a screenshot dump. Six surfaces, in one document, that add up to "here is how the month went and what to do next."
  • Cover the whole funnel: AI visibility (by engine), keyword rankings, website traffic, Search Console, local presence, and competitive standing with next moves.
  • Every proportion carries n and a 95% confidence interval. That is how a client tells a real move from sampling noise.
  • Never fake what you cannot measure. Paid media and private Google Business Profile clicks come from owner-only sources. Say so, do not estimate.
6
surfaces folded into one client-ready report
95%
confidence interval on every proportion
1
cover page you can white-label per client

Most "monthly reports" are five tool exports stapled together: a rankings CSV, a traffic screenshot, a slide of AI mentions, and two charts nobody reads. It is a pile of numbers, not a report. A report answers a question the client actually has: did this month go well, and what should we do next?

Answering that takes a deliberate set of surfaces, in one document, framed honestly. Here is what belongs in a monthly AI search report, and why.

llemmy · Monthly Performance report
One report, the whole picturecover page · white-label ready
AI Visibility
58%
95% CI 51–65% · n=180
Share of Voice
34%
95% CI 30–38% · n=402
In top 3
41
keywords ranking 1–3
In the pack
75%
local searches · 95% CI 47–91%
AI search performancevisibility over time, by engine
Keyword rankingsmovers + most searched
Website trafficGA4 + AI referrals
Search Consoleclicks, impressions, queries
Local presencepack rank, rating, reviews
Competitive standing + next moveswho leads, what to do

The llemmy Monthly Performance report folds six surfaces into one branded, cover-paged document. Illustrative data.

1. AI visibility, engine by engine

Start where search is going, not where it was. Show how often the AI engines name your brand when someone asks about your category, and show it per engine, because ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews do not agree. A single blended number hides the engine you are losing. Track it over time so the client sees the trend, and put a confidence interval on it so a two-point wobble does not get sold as a win.

2. Keyword rankings, prioritized

Organic rank still feeds both classic search and the AI answers built on top of it. But a client does not need all 214 tracked keywords in a wall of rows. They need the movers (what climbed, what slipped, what newly ranked) and the most searched. Rank is a point estimate from a single daily sample of a personalized, localized SERP, so present it as directional, and reserve confidence intervals for genuine proportions like "share of keywords in the top 3."

3. Website traffic, including AI referrals

Visibility is the top of the funnel; traffic is the middle. Pull the analytics that matter (new users, engagement, the channels people arrive through) and separate out the visits that came from an AI assistant. That AI-referred line is small today and growing, and it is the clearest evidence that being cited in answers turns into real people on the site.

4. Search Console

The classic-search complement to the AI story: the searches that put your pages in front of people, how often they clicked, and where you rank on average. It grounds the report in first-party Google data and gives the demand context that AI visibility alone cannot.

5. Local presence, for local businesses

If the client serves a place, the map pack is where nearby customers decide. Include the local-pack rank for the searches that bring them in, plus the star rating and review count that shape whether anyone clicks. Like organic rank, pack position is a daily point estimate; the honest headline is the share of local searches where you appear in the pack, which is a proportion and carries a confidence interval.

6. Competitive standing and the next three moves

End with orientation and action. Show where the brand sits against its rivals across the surfaces above, then close with a short, specific list of what to do next month. A report that ends in "here is the number" is a status update. A report that ends in "here is the move" is worth paying for.

The rule that makes it a report, not a brag

Two disciplines separate a trustworthy report from a flattering one.

Show your uncertainty. AI answers are sampled and change on their own, so every rate is a point on a range. A visibility of 58% is only meaningful next to its sample size and 95% confidence interval (51 to 65%). That is what lets a client read a month-over-month change as real or as noise, instead of taking every uptick as a victory.

Never claim what you cannot measure.

Paid media spend and private Google Business Profile metrics (call clicks, direction requests, website clicks) come from owner-authenticated accounts, not from reading public search. If your tooling cannot see them, name them as out of scope. A report that is honest about its own edges is one a client keeps trusting; one that quietly estimates the parts it cannot see is one they eventually catch.

Build the month around those six surfaces, frame every number with its uncertainty, and be clear about what you did not measure. That is a report a client reads to the end, and forwards to their boss. llemmy's Monthly Performance report assembles all six into one branded, cover-paged document, and leaves out any section you have not connected, so it always reads clean.

See how AI describes your brand

Run a free GEO audit โ€” no signup needed to see your score โ€” or start tracking your brand across every AI engine.