AI visibility
AI visibility is a brand's measurable share in AI assistant answers: how often it is mentioned, whether it is cited as a source and in what wording it is described. It is not one metric and no tool measures it directly. It is assembled from several indirect indicators.
In short
| What it is | How often and how accurately a brand appears in assistant answers |
| What it consists of | Citations in assistants, impressions on question-shaped queries, sessions from assistant domains, accuracy of the brand description |
| What influences it most | Crawler availability, original data in the content, consistent facts across sources |
| What the real numbers look like | Single-digit sessions per month at the start. Watch the trend, not the volume |
| Common misreading | Comparing it to organic traffic. Different order of magnitude, different purpose |
How it is measured
From four sources in practice. First, citations in Bing Webmaster Tools, which has an AI Performance report. Second, Search Console filtered to queries longer than six words or containing a question mark, the closest proxy for conversational queries. Third, analytics, where sessions from assistant domains are tracked. Fourth, a manual test of a sample of queries, recording whether the brand is mentioned and how.
None of them is sufficient alone. Together they produce a trend you can follow on a single row per month.
From our own practice: the numbers are small at the start
Visibility in AI starts at zero in practically every case, and the first sessions from assistant domains arrive only after months of work on content and availability. By default analytics files them among ordinary referrals, so the first step is to give them a channel of their own. Without it there is nothing to track.
The citation baseline comes from the AI Performance report in Bing Webmaster Tools, the only tool that reports AI citations directly from the operator, and it is tracked monthly. The third number is availability: in an audit the crawlers are tested with direct requests and every one that gets HTTP 403 because of the firewall, not because of robots.txt, is fixed.
That is the realistic order of magnitude at the start. Anyone expecting hundreds of sessions will be disappointed. The value of the channel is that the brand is mentioned at the moment of a decision.
Common mistakes
- Demanding a single number. No unified metric exists, and tools that promise one are reporting their own estimate.
- Measuring before there is anything to measure. If crawlers have no access, visibility never forms and the report stays empty.
- Judging it after a month. The first movements take months, not weeks.
- Ignoring what the model says about the brand. An inaccurate description matters as much as the number of mentions.
Related terms
See also GEO, AEO, AI Mode, AI Overviews, RAG and llms.txt.
Frequently asked questions
Is there a tool that measures this for me?
Several offer it, but all of them work from their own sample of queries. They are usable for a trend, not as an absolute figure.
How often should it be measured?
Monthly is enough. Measuring more often only adds noise, because the numbers are small.
Can AI visibility be bought?
No. Citations, no. Paid ads do appear in AI answers (AI Overviews, AI Mode), but they are labeled as ads and do not earn citations; those the model selects on content and availability.
How we can help
AI visibility is a standalone service for us, measurement and availability audits included. Details are on the AI visibility agency page.