Visibility score

The visibility score sums up, in one number from 0 to 100, how present your brand is in the answers AI engines give to your questions. It is built only from answers Rixe actually collected, it always comes with the sample behind it, and rates built on too few answers are marked as such instead of being shown as confident numbers. Use it to follow the trend and compare yourself with competitors; use the per-prompt and per-engine views to find out why it moves.

#Where to find it

Go to AEO Insights › Visibility (in the Analytics group of the sidebar). The tab shows:

• AI visibility — the score on a dial, with a sentence such as "When buyers ask AI about businesses like yours, you appear in 12 of 100 answers."
• A status line under the page title — your score, your rank among the tracked brands, and the sample: "1,240 answers · 48 prompts · 3 engines · last 30 days".
• Visibility — the trend chart, one line per brand. A day a brand was not measured is left as a gap rather than joined with a straight line.
• Competitor ranking — the brands with the highest visibility on your questions, with share of voice, sentiment and position.
• Position distribution, Mentions & citations, Decision journey and Do the engines say the same thing? — the detail behind the number.

#What goes into the score

For each engine, Rixe combines three measurements of your brand over the period:

The per-engine results are then averaged using your market's engine weights, so an engine that few people in your country use does not swing your score. The result is scaled to 0–100.

Two rules keep the score honest:

• Only search-grounded engines count. ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Microsoft Copilot and Grok search the web before answering. Chat models without search (DeepSeek, Mistral, Claude, Qwen) answer from training data and rarely name a niche or local brand however well it ranks, so they are measured and shown per engine but carry no weight in the score.
• An unweighted engine is not dropped. If an engine was measured but your market pack has no weight for it, it is counted at a default weight and the page says so ("These engines were measured but carry no weight in the market pack, so a default weight was used").

#Reading the dial

The dial is split into five bands:

#The confidence interval

Every rate comes with a range, not just a point: a Wilson confidence interval at 95% confidence. You see it as text such as "Confidence interval 18%–34% · 52 runs" next to a rate, and in the client portal and PDF reports as a shaded band around the trend line.

• A narrow band means many answers stand behind the number.
• A wide band means few do — the true value could be anywhere inside it.
• When two brands' or two periods' bands overlap, the difference between them is not statistically meaningful, however different the midpoints look.

#Not enough data

A rate based on fewer than 4 runs is not shown as a number. You see Not enough data (some screens say insufficient sample or Only 3 runs — this rate is not reliable) instead. This clears on its own as daily measurement adds answers.

#Filters

The strip above the tabs applies to the whole view and stays in place when you switch tabs:

1. Period — the last 7, 30 (default) or 90 days.
2. Brand — view the page as a competitor would see it (competitor simulation).
3. Unbranded — recalculate from category prompts only, leaving out prompts that name your brand. This answers a narrower question: when the engine is not led by a name, who does it reach for?
4. Engines, topics, persona and market — narrow the data to a subset.

#Why the score moves

Your score can change without you doing anything, because the engines change: models are updated, indexes refresh, and a page cited last week may not be cited this week. Rixe helps you tell the difference:

• Significant changes only. Rixe flags a change as real only when both periods have at least 4 runs, the difference is statistically significant (p < 0.05) and it is at least 10 percentage points. Significant changes can trigger alerts (Alerts and inbox).
• Engine breaks. When every brand measured on an engine falls or rises on the same day, the chart marks a break and says so: "this drop did not come from anything you did". When only your brand moved, it says that instead.
• Degraded engines. If an engine is having problems, a banner names it and its missing answers are left out — your score is not lowered because of a failure.
• Engine agreement. Do the engines say the same thing? shows where the engines disagree on a question. Where they disagree the answer is unstable, and that is where a change moves the most.

#Common questions

Why is my score different from my mention rate? The score also counts citations and position, weights engines by your market and leaves out chat models without search. The mention rate is just named ÷ answers.

Can I export the numbers? Yes, the strip has an export for this tab. For automated access, see the API.

#Related pages

• How measurement works — runs, samples and missing data.
• Concepts — share of voice, citation, position and other terms.
• Competitors — comparing yourself with rival brands.
• Mentions and perception — what the answers say about you.