Measurement
What is an AI visibility score — and how is it calculated?
Updated August 2026
An AI visibility score is a single 0–100 number summarizing how present your brand is when AI assistants answer your buyers' questions. Behind the number sit a handful of measurable components — and knowing them tells you exactly which lever to pull when the score is low.
The components that make up the score
A credible score is computed from repeated, identical questions asked across AI engines, then aggregated. These are the components Mentioned measures — most serious methodologies use some weighting of the same five:
| Component | What it measures | How to move it |
|---|---|---|
| Mention frequency | The share of AI answers that name your brand at all | Publish pages that answer lost questions verbatim; get listed in the sources AI cites |
| Position | How early you appear when answers list options (1st vs 7th) | Strengthen the evidence for your primary use case — models lead with the option they're most confident about |
| Competitor comparison | Your mention share relative to the most-mentioned competitor | Comparison and alternatives pages; win the questions where the gap is smallest first |
| Citation presence | Whether your own site appears among the sources answers cite | Crawlable, direct-answer content; allow AI crawlers; schema markup and llms.txt |
| Recommendation confidence | Whether you're recommended affirmatively or mentioned with hedging | Entity consistency — one clear description of what you are, everywhere |
What a good score looks like
Honest benchmarks from scanning real brands:
- 0–30: where most non-famous brands start — absent from most relevant answers. Normal, and fastest to improve.
- 30–70: AI engines know you and name you for a meaningful share of questions. The work shifts from “exist” to “win specific questions.”
- 70+: category-leader territory — named in most relevant answers, usually early. Defense matters as much as offense here.
The trend beats the level. A brand moving 20→40 after publishing three comparison pages has proof its GEO work lands; a static 55 tells you nothing new.
Why scores move when you change nothing
Two reasons. First, variance: the same question can yield slightly different brand lists run to run, so single-scan movement of a few points is noise — read direction across several scans. Second, the denominator is alive: engines refresh retrieval sources continuously, and competitors publish too. A falling score with no change on your side usually means someone else closed a gap you were winning by default.
How to actually raise it
Work question-by-question, not score-by-score. Sort your tracked questions into: winning (defend), close (competitor named, you sometimes present — highest ROI), and absent (needs new content or new sources). Ship one targeted fix per question — a direct-answer page, a comparison page, a listing on a source AI already cites — then re-scan and check that specific question. The score follows the questions; chasing the aggregate directly leads to unfocused work. The full tactical list is in how to get recommended by ChatGPT.
Frequently asked questions
What is a good AI visibility score?
Most brands start below 30. 40–70 is solid presence; 70+ is category-leader territory. Trend beats level.
Why did my score change when I changed nothing?
Run-to-run variance plus live engines and active competitors. Judge direction over several scans, not one.
Are scores comparable across tools?
No — there's no standard formula. Compare only within one tool and method over time, and prefer tools that disclose their components.
Related: AI visibility monitoring guide · GEO vs SEO