Cleotic

Visibility metrics

On this page

Cleotic's visibility figures describe the answers it captures for your tracked questions and models. They are not a share of every conversation on an AI platform, and they are not market share.

Visibility

Visibility is a brand's mention rate: the percentage of captured answers in which Cleotic detects the brand. If your brand appears in 15 of 20 answers, its visibility is 75%. One question can produce several answers across models and scheduled runs.

Average position describes how early the brand appears among the detected brand mentions. Lower positions mean earlier mentions. It is calculated from answers with a detected mention and is shown separately; it does not change the visibility figure.

Visibility does not measure whether the mention was favourable. Read Perception and the answers alongside it.

Share of voice

Share of voice is a brand's percentage of the answers that name your primary brand or an active competitor, within the selected scope.

Each answer counts once for each tracked brand it names, however many times it repeats the name. If 45 answers name your brand and the tracked brands are named 100 times across answers in total, your share of voice is 45%.

Visibility uses captured answers as its denominator; share of voice uses the brand appearances across those answers. Read them together: you can have a large share of a very small number of answers. Adding, pausing or deleting a competitor changes the denominator.

The likely range

AI answers vary from run to run, even for the same question. Each rate on Overview therefore shows the range it likely sits in, for example Likely 28–41%. The range is a 95% interval built from how much each of your questions varies from run to run. It narrows as more answers arrive, and it is tightest when each question reliably does or does not name the brand.

A change between periods counts only when it is larger than that run-to-run variation could explain. Otherwise Cleotic shows No real change on the headline figures and leaves out the arrow elsewhere. Brands within normal variation of each other share a place, for example Level with Acme for #1.

The range describes your tracked questions as they are. Adding or removing questions changes what is measured, so compare periods with the same question set.

Calibrating

A rate resting on fewer than 10 answers is calibrating. Share of voice calibrates until 10 answers name a tracked brand. Cleotic shows the figure with a note, because it can still move a lot as answers come in. New brands, new segments and new engines usually calibrate within their first scheduled runs.

The trend chart shows how visibility moves over time, as a percentage. Hover over a point for its value and use the legend to identify each brand. All brands on Overview puts every competitor on the chart.

For periods of three weeks or more, Overview plots one point a week across the previous period and this one, with a dashed line where this period starts. The shaded band around your brand's line is the range each week likely sits in. Weeks with few answers have wider bands. Shorter periods plot one point a day.

Longer windows show sustained patterns and give narrower ranges. A long date selection cannot reveal history outside your plan's retention window.

By model

The per-model breakdown shows each model's mention rate for a brand, the same measure as visibility, with its own likely range. One model has fewer answers than all models together, so its range is wider.

Google AI Overviews and Google AI Mode appear as separate entries from Gemini when your segment includes them. Use the model names shown in the app; availability depends on your plan and the segment.

A stronger result on one model does not tell you what was in its training data. Differences may come from response variation, model behaviour, available grounding, or the questions sampled. Don't assume that a content change will improve every model together.

Compare periods fairly

Choose the date range and segment before comparing. Then check:

  • The same prompts and models contributed answers in both periods.
  • Collection didn't pause and model runs didn't fail.
  • Prompt wording, brand aliases and active competitors didn't change. Alias and competitor changes start reanalysis, which can change historical counts.
  • The move is larger than the likely range. A line that stays inside its band is within normal variation.

A real change says the numbers moved, not why. Open the answers behind a change, and the model breakdown to see whether it is concentrated in one model, before crediting it to your own work.

Track a content change

  1. Record the publication date and the questions the content addresses.
  2. Keep those prompts and their segment consistent.
  3. Compare several later runs with the earlier baseline.
  4. Review citations and the actual wording as well as the figures.
  5. Share the evidence and its date range in a report.

Content Studio measures each page it publishes in its own results. There is no universal target score: use your own baseline and relevant competitors.

Updated