AI Visibility Tracking: What You Can Measure and What You Cannot

Traditional search had a measurement culture: positions, impressions, clicks. AI-generated answers do not fit neatly into it.

There is no ranking. An AI answer is a composed response, perhaps with a few cited sources. Being mentioned is not the same as being "position 3".

Answers vary. The same question can produce different answers on different days, for different users, in different phrasing, with different model versions. A single check is an anecdote.

Platforms report little. Google Search Console includes AI feature traffic within overall Search performance and does not break it out as a separate "AI Overviews" report you can use to track mentions, and AI assistants provide no webmaster console at all.

Vendors fill the gap with prompts. Many "AI visibility" tools sample a library of prompts across assistants and report how often your brand appears. That is a reasonable research technique, but it is a sample of a sample, and a score of 37 percent is only meaningful relative to the same method over time. It is also easy to inflate by choosing favourable prompts.

Given that, the useful question is not "what is my AI ranking?" It is "am I doing the things that make me easy to find, understand and trust, and are my signals improving?" That is something you can measure.

1. AEO readiness over time

Each AEO analysis stores a score and the underlying signals (structured data, FAQ and question coverage, trust and authority signals, entity clarity). Running it after each batch of improvements gives you a trend line: the site is measurably clearer and better-documented than it was three months ago. See the AEO analysis page for how scores are produced and their limits.

2. Web brand mentions

Answer engines draw on what the web says about you, not only what you say about yourself. Brand mention discovery searches the web for pages that mention your brand without linking to your site, stores them, and lets you generate a polite outreach draft asking for a link. Beyond link building, mentions are an indicator of how often and where your brand is discussed.

3. Search presence

Rank tracking records your organic position, your Google Maps position for local keywords, and which SERP features appear. Because AI features in Google Search surface alongside ordinary results and draw on indexed pages, strong traditional visibility remains the base.

4. Real traffic from Google

Connect Google Search Console and Google Analytics to see impressions, clicks, queries and landing pages. This is the only first-party data about how people actually reach you from Google, including from its AI features, which are reported within overall Search performance.

5. Your own spot checks

We recommend a small, documented manual routine: pick 10 to 20 questions your customers ask, ask them in the AI products they use, and record whether and how you are mentioned. Do it monthly with the same questions, noting the date and product. It is crude and it is honest.

No single signal is "AI visibility". Together they tell a story.

Set aside an hour each month. The aim is a consistent record, not an impressive dashboard.

A local services company

A landscaping firm asks five AI products ten questions about hiring a landscaper in its city. It is mentioned in two of fifty responses. Reviewing the cited sources, it notices that the competitors named most often have detailed pricing guides and project galleries with named locations. It writes its own pricing page, adds service-area pages, runs an AEO analysis to catch missing structured data and FAQs, and re-checks in three months. The result is not guaranteed; the process, though, is concrete and documented.

A SaaS company

A software company finds that its brand is rarely mentioned on third-party sites. Brand mention discovery surfaces eight articles that mention the product without linking. Outreach converts three into links, and the team starts contributing expert commentary to two industry publications. Mentions grow, and the log of spot checks shows slightly more frequent appearances for comparison questions.

An agency reporting to clients

The agency cannot promise AI citations, so it reports on what it controls: readiness score movement, number of questions answered on-site, mentions gained, and Search Console impressions. Clients receive an honest report that connects activity to measurable signals.

Agencies that need a defensible way to report on "AI readiness" without overpromising.

Brands with strong offline reputations who want to know whether the web reflects that reputation well enough for answer engines to pick it up.

Content teams looking for evidence of which questions their site answers well and which it does not.

If you specifically need automated prompt-sampling across AI assistants today, this feature does not do that, and we would rather say so than imply otherwise.

The two work together. Automation handles the measurable on-site and web signals. Manual checks cover the part that cannot be reliably automated.

Benefits

• No made-up numbers: every signal has a clear definition.
• Focuses effort on changes that help people and search engines alike.
• Gives agencies something truthful to report.

Limitations

• It does not measure AI citations directly. See the status box at the top of this page.
• Readiness is not outcome. A better readiness score is a reason to expect clearer content, not a promise of mentions.
• Brand mention discovery depends on the web search source configured for the platform and finds what that source returns.
• Manual spot checks are subjective and labour-intensive; keep the question set small and consistent.

1. Keep the question set fixed and dated.
2. Record sources cited, not just whether you were mentioned.
3. Compare against two or three real competitors.
4. Treat improvements as hypotheses and look for corroboration in Search Console.
5. Prefer improving the page that is cited for competitors to creating new thin pages.
6. Do not buy mentions. Earn them with useful content, data and expertise.

• Trusting a single AI answer. One response proves little.
• Optimising prompts instead of pages. You cannot control what customers type.
• Buying "AI SEO" packages that promise citations. Nobody can guarantee them.
• Neglecting normal SEO. AI features in Google Search build on indexed, helpful pages.
• Stuffing pages with FAQ schema. It is not a shortcut.