AI visibility
Updated 2026-09-15
AI visibility is the degree to which a brand is mentioned, cited or recommended inside answers generated by AI systems such as ChatGPT, Perplexity, Gemini and Google AI Overviews. It is a sampled rate rather than a position, because the same question can produce different answers on two consecutive runs.
How it is calculated
There is no single "AI visibility score" that means anything on its own. In practice the term is an umbrella over four measured rates — mention rate, prompt coverage, share of voice and citation rate — each computed over answers that were actually grounded in a search, with failures and missing surfaces excluded rather than counted as zeros. Any vendor reporting one composite number should be able to show you the four underlying rates and the sample size behind them.
Worked example (illustrative numbers, not a customer measurement)
A brand runs 30 prompts, each sampled 3 times on 3 engines: 270 answers. If 61 of them mention the brand, mention rate is 61 / 270 ≈ 0.226. If the brand appears at least once for 9 of the 30 prompts, prompt coverage is 0.30 — a useful reminder that a respectable mention rate can hide the fact that two thirds of the prompt set never surfaces the brand at all.
Common mistakes
- Treating one manual ChatGPT check as a measurement. A single answer is a coin flip, not a rate.
- Counting a missing AI Overview as "we were left out". If Google generated no overview, there was nothing to be left out of.
- Comparing two tools' "visibility scores" directly. Different prompt sets and sample sizes produce different numbers for the same brand.
Frequently asked questions
- Is AI visibility the same as SEO ranking?
- No. A ranking is a position in a list that you can capture with one crawl; AI visibility is a sampled rate of appearing inside a generated answer, where no fixed position exists and repeated measurement is the only way to get a stable number.
- What is a good AI visibility number?
- There is no category benchmark worth quoting, because the number depends entirely on your prompt set, your competitors and the engines you measure. The useful comparison is your own trend across cycles with the prompt set held constant.
