How to measure AI visibility

Choose a repeatable set of questions and connect each metric to the answers behind it.

By the chatobserver teamUpdated: September 6, 2026

Choose the questions people ask about your category, the platforms they use, and the markets you serve. Keep that scope visible beside every result so a change in the sample is not mistaken for a change in performance.

For example, if your brand appears in 64 of 140 collected answers, its answer-level visibility is 45.7%. If the previous period contained 52 mentions in the same number of answers, the increase is 8.6 percentage points. Those figures describe this sample, not every conversation on the platform.

See what the denominator changes.

Count each answer once when it mentions the brand. In chatobserver, this answer-based presence rate is called share of voice.

Share of voice45.7%
Answers analyzed140
This is a sample rate, not a measure of every AI conversation.

Worked example

Start with

Choose ten questions in one category, one language, and one market. Collect them on two platforms over seven days. If every planned answer completes, that creates 140 answers. Keep a separate record of planned runs and failures; a missing answer must not silently become a brand absence.

Read the evidence

Suppose 64 of the 140 collected answers mention Klaviyo. Its share of voice in this sample is 64 ÷ 140 × 100 = 45.7%. Repeating the name five times in one answer still counts as one answer mentioning the brand. Two brands can both be present, so their percentages need not add to 100%. Preserve platform-level counts: a blended result can move merely because the sample mix changed.

Make the decision

Compare this with 52 of 140 answers from the previous period: 37.1%, a rise of 8.6 percentage points. That is an observed difference, not evidence that a content edit caused it. Confirm the question set, platform mix, dates, and completion coverage match before interpreting the direction.

Take this away

Publish the numerator and denominator next to the percentage. Keep new questions in a separate group until you have a comparable baseline. Inspect a few changed answers to explain the result in words.

Reusable outline

Download text

Define the sample before the score

  • Choose a fixed prompt set and keep intent labels attached to it.
  • Track mention share, citation rate, and competitor coverage together.
  • Review answers by assistant and by market rather than only in aggregate.
  • Report deltas over time instead of one-off wins.

Review changes in context

  1. Establish a weekly review rhythm.
  2. Separate educational and commercial prompt performance.
  3. Flag prompts where the brand is absent or weakly cited.