Skip to contentNewStarter $24, Growth $48, Business $90, articles included · annual: 2 months free
All articles

How agencies report AI search visibility to clients

Citable · 20 October 2026 · 7 min read

Agencies report AI search visibility by tracking a fixed set of buyer prompts across each AI engine and showing a short set of metrics: visibility, position, sentiment, citations and share of voice, each with its formula and denominator. Monthly reports work best when the numbers are sampled over several weeks and paired with a log of the actions taken, so a change can be checked against what moved. Reports should state which prompts and assistants were included and leave out unavailable data rather than scoring it as zero.

  • Keep one short metric set and use it for every client and every month.
  • Put the formula and the denominator next to each number.
  • Sample over several weeks, and act only on shifts that hold for three to four weeks.
  • Log each action, then report whether the metric moved, stayed flat or fell.

Which metrics belong in a client report?

AgencyAnalytics recommends tracking four metrics per AI engine (visibility, position, sentiment and citations) against a fixed set of prompts, reported monthly.

Illustration about “Which metrics belong in a client report?”

SearchAtlas lists five attributes for the same job: mention rate, citation rate, share of voice, sentiment and position, sampled from a fixed prompt set. The lists overlap but differ. What matters is choosing one and using it for every client, because changing the set between months makes the trend unreadable.

Report per platform, not as one blended score. ChatGPT, Gemini, Perplexity, Claude, Grok and Google AI can treat the same brand differently, and an average hides the differences the client needs to see.

Why does each metric need a definition?

The same week of answers can produce very different share-of-voice figures. GrowByData describes a case where one slide said 40%, another 20% and the citation report 6%.

The cause is the denominator. Mention rate, mention share and citation share count different things, so each needs its own label. A client who sees "share of voice: 20%" should be able to tell which of the three it is.

Put the formula and the denominator on every slide. It takes one line per metric and removes most follow-up questions.

Count mentions and citations separately and show them side by side. GrowByData recommends reporting mention share and citation share together because they move for different reasons: a brand can be named often and linked rarely.

How do you keep months comparable?

A monthly trend only means something if the inputs stay the same. Some agency guides say the platforms, questions, run count and schedule should be frozen for a period. SearchAtlas describes freezing all four for the quarter, so each month compares with the last.

Write the prompt set down and send it with the report. If you add or remove a prompt, say so and mark the break in the trend.

Sample over several weeks before you report. AgencyAnalytics advises sampling over weeks and reporting monthly, so you don't walk a client through a drop that was not real.

The sources themselves churn. Semrush's AI Visibility Index work, reported by Search Engine Land, found that 40 to 60% of cited sources change from month to month. A single-month citation comparison is fragile; a multi-week average holds up better.

How do you show what changed?

A useful report shows the work as well as the numbers: a change log of agency actions placed next to the metric each one was meant to affect. SearchAtlas pairs each metric with such a log.

Treat a shift as a signal only when it holds. GrowByData's rule of thumb is that a share-of-voice shift lasting three to four weeks is worth acting on, since week-to-week swings are normal.

Then close the loop. Log the action, wait for the sampling window, and report whether the metric moved, stayed flat or fell. That step tells the client whether the work did anything.

Can you prove ROI from AI visibility?

Agency guides disagree. AgencyAnalytics says that putting AI visibility in the same report as traffic, conversions and campaign performance lets you prove ROI.

SearchAtlas takes the more cautious line:

"The report describes observed visibility and never claims that a citation produced revenue." — SearchAtlas

Place visibility and traffic in the same report, but present them as separate series unless you can show the link between them. The defensible version reads: visibility moved, here is the work behind it, and here is what the data does not show.

What should a report state, and how often should you send it?

A transparent report states which buyer questions and assistants were included. GetGeoPulse also says to omit unavailable data instead of scoring it as zero. A zero and a missing value tell the client different things.

Cadence is where guides split. Demand Local proposes four tiers; AgencyAnalytics advises monthly reporting only.

CadenceProposed by
Daily anomaly alertsDemand Local
Weekly digestDemand Local
Monthly white-label PDFDemand Local
Quarterly reviewDemand Local
Monthly reportAgencyAnalytics

Both are defensible. More frequent updates react faster but show more routine movement. Escalate an anomaly only when it is sustained or tied to a known change.

Why report third-party mentions too?

A report that tracks only the client's domain misses most of the picture.

Split the report into owned-site work (page fixes, structured data, articles) and off-site work (mentions, links, outreach). They respond to different actions, so track them in separate rows. Keep the source list visible: when a mention comes from a directory, a review site or a press page, the client should see which one, because that is where the next action goes.

What does branded reporting cost?

The reporting layer has a price. Ayzeo includes white-label PDF and DOCX reports in its $149/month Pro plan and sells a separate White-Label Dashboard add-on.

Decide whether the client pays for the report or the agency absorbs it. Price the tracking separately from the work that changes the numbers, so the client can see what each one buys. For vocabulary, Citable's explainer on what GEO and AEO are is useful additional reading.

Next steps

  • Choose one metric list (four or five metrics) and write the formula and denominator for each.
  • Freeze the platforms, prompts, run count and schedule for a quarter, and save the prompt set in a shared document.
  • Start a change log with a date, the action and the metric it targets, and review it each month.
  • Add a "not available" label to your report template so missing data is never shown as zero.
  • If you want a first snapshot of whether ChatGPT mentions a site, Citable's free AI visibility report shows it, and Citable also offers reports and an API for agencies.

Frequently asked questions

How often should AI visibility reports be sent to clients?

AgencyAnalytics advises sampling over several weeks and reporting monthly. Demand Local suggests adding daily alerts, a weekly digest and a quarterly review.

What is the difference between mention rate and citation rate?

Mention rate counts how often a brand is named in answers. Citation rate counts how often the brand's pages are cited as a source. GrowByData says to report them separately because they move for different reasons.

Can you show a client revenue from AI search visibility?

Guides disagree. AgencyAnalytics says combining visibility with traffic and conversions lets you prove ROI, while SearchAtlas says a report should describe observed visibility and never claim a citation produced revenue.

Should AI visibility be reported alongside SEO rankings and traffic?

Yes, in the same report, as separate series. That gives the client context without implying that visibility caused the traffic or conversions beside it.

Sources

Keep reading

Want this measured for you?

Get a free report on up to five buyer questions.

Get the free report