An AI Visibility Report Template for Agencies and In-House Teams

What the report has to do

An AI visibility report answers three questions for someone who doesn’t read dashboards: Are we being recommended? Who is recommended instead? What are we doing about it? Everything below serves one of those three. Keep it to one page plus an appendix.

Section 1 — The headline (three numbers)

Three tiles at the top, each with last period’s figure beside it and the sample size underneath:

Always show counts (“9 of 30”) alongside percentages. It stops a client reading a move from 1 of 5 to 2 of 5 as “visibility doubled”. Our formula guide has exact definitions to paste into the appendix.

Section 2 — By model

One row per assistant. This is usually the most actionable table, because models disagree.

Section 3 — Who is recommended instead

The five brands named most often in the same answers, with their mention counts. Clients care about this table more than any other: it turns an abstract metric into a named rival.

Section 4 — The questions we’re losing

The three to five buyer questions with the lowest visibility, each with one short quote from a real answer showing who was named. Quotes make the problem concrete and are hard to argue with.

Section 5 — Where AI gets its information

The domains most often cited in the answers: review sites, directories, publications, Reddit, YouTube. This is the to-do list for earned media, and it shows the client why PR and reviews belong in an “SEO” budget.

Section 6 — What we did, and what’s next

Two short lists. Last period’s actions (content published, listings fixed, mentions earned, technical changes such as crawler access or llms.txt), and next period’s, each tied to a losing question or a cited source above.

Section 7 — Method (appendix)

One paragraph the client can forward to a sceptical colleague:

Method: We asked [N] buyer questions, [R] times each, to [models]
between [dates] — [total] answers in all. A brand counts as named when
its name appears as a whole word in the answer. Visibility = answers
naming the brand ÷ all answers. Share of voice = the brand's mentions ÷
all brand mentions. With [total] answers, a visibility reading carries a
margin of error of roughly ±[x] points, so changes smaller than that
are not treated as real.

Look up the ± figure for your sample size in our margin-of-error tables.

Wording that keeps you honest

• Say “named in 30% of sampled answers”, not “ranks #1 in ChatGPT”. AI answers have no fixed ranking.
• Say “within the margin of error” for small moves, and resist explaining them.
• Don’t promise placements. Nobody outside the model companies controls what a model says; you can improve the signals and measure the outcome.
• Keep the questions frozen between periods, or the trend line measures your question changes rather than the market.

Where the numbers come from

You can fill every table by hand from spot-checks, or with any tracking tool. Mentioned produces sections 1–5 for each tracked keyword, exports them as CSV, and keeps the questions frozen between runs. A free check gives you a first snapshot for a new client.

Run a free check