
You can rank on page one for the search that matters most to your business and still lose the visitor. The answer may already be on the screen, written by an AI assistant that cited someone else.
The short answer: to measure AI search visibility, track three signals that rankings do not show. First, whether AI assistants such as ChatGPT cite your website when asked the questions your buyers ask. Second, how many visits arrive from AI assistants, and on which pages. Third, whether a Google AI Overview appears above your result on your most important searches. All three can be checked by hand in about an hour a month.
This guide is for marketing leads and founders of small and mid-size business websites who want evidence, not impressions, of whether AI search sends people their way.
A ranking report says where your page sits in Google's list. That was a good proxy for "do people find us" when the list was the whole result. It is weaker now, for three reasons.
Rankings still matter, but they describe one surface out of several.
A single "AI visibility score" can blend signals you cannot check yourself. Three signals can be observed directly.
| Signal | The question it answers | What it cannot tell you |
|---|---|---|
| Cited in AI answers | When buyers ask an assistant about your topic, does the answer link to you, name you, or point to a competitor? | How many real people asked that question |
| Visits from AI assistants | Are assistants actually sending people to your site, and to which pages? | Visits where the assistant passed no referrer |
| Appearing under AI Overviews | On your most important Google searches, is there an AI answer above your result? | Clicks from the AI Overview on its own, which Search Console does not report separately |
You need all three: a site can be cited often and receive few visits, or receive visits on pages nobody thought of as AI-friendly.
A citation is a link to your site inside an AI answer; a mention names your brand without a link. Track both: they need different fixes.
Pick ten to twenty questions a real buyer would ask, in their words. Good sources are the searches your pages already appear for in Search Console and the "People also ask" questions Google shows for them. Include comparison and problem questions. Leave out your own brand name: an assistant is likely to cite you for it, and that hides the real picture.
Then fix the list. Change it monthly and you measure your choice of questions, not your visibility.
Ask each question in ChatGPT with web search on, so the answer shows its sources, and ask it more than once in a fresh conversation: one answer is one sample, and a handful can differ a lot. For each answer, record whether your site is cited, whether your brand is named, and which other sites are cited.
A figure such as "cited in 4 of 12 answers" is useful; "we showed up" is not. The share of answers citing you, and the sites cited instead, is your baseline. With a dozen answers, a change of one or two between months is noise.
Google Analytics 4 has a default channel for this. Google defines the AI Assistant channel as "the channel by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok", and assigns it when the referrer matches Google's list of AI assistants (Google, "Default channel group").
To read it:
When you read the numbers:
On eligibility, Google is direct: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." A page must be indexed and eligible to be shown with a snippet (Google Search Central). So if a key page is missing from AI Overviews, check the basics first: indexing, and whether a nosnippet or max-snippet setting is limiting what Google can show.
| What to measure | Where to find it | How often |
|---|---|---|
| Share of answers that cite you, name you, or cite someone else instead | ChatGPT with web search, your fixed question list, several runs per question | Monthly |
| Visits from AI assistants, by assistant | GA4 traffic acquisition: AI Assistant channel, then session source | Monthly; weekly if it is growing fast |
| Pages AI assistants send people to | GA4: AI Assistant channel with landing page | Monthly |
| AI-assistant landing pages that are broken | GA4 landing pages checked against your not-found page | Monthly |
| AI Overviews above your key searches | A private-window Google search for your top non-brand searches | Monthly |
| Clicks and click-through on those searches | Search Console Performance report, Web search type | Monthly, against the previous period |
| Crawler access and snippet settings | robots.txt, page meta robots tags | After any robots.txt or template change |
Each signal points to a different kind of fix.
Then measure again against the same questions and searches; without a fixed baseline you cannot tell a fix from a fluctuation.
The manual routine works, but the question list drifts, a dozen answers is a small sample, and comparing spreadsheet tabs is slow.
Website Growth Ops, a Tomorrow Central software, automates the measurable parts of this routine for a website you have verified:
It has read-only access to Search Console and Analytics and never changes your website; you make the change. It asks one assistant, ChatGPT, covers Google Search and GA4 only, and is switched on per organisation while it is new.
Track three signals: whether AI assistants cite your site for the questions your buyers ask, how many visits arrive from AI assistants and where they land, and whether AI Overviews appear above your result on your most important Google searches. Check each one monthly against the same baseline, so a change means something.
Ask ChatGPT, with web search switched on, a fixed set of questions your buyers ask, several times each, and record whether your site is cited, whether it is named, and which sites are cited instead. Keep the same questions month to month, or the trend will not mean anything.
In GA4's traffic acquisition report, look for the AI Assistant default channel, which Google describes as visits from AI assistants such as ChatGPT and which excludes AI Overviews and AI Mode. Then look at session source and landing page to see which assistant sent the visitors and which pages they arrived on.
Google says AI Overviews and AI Mode traffic is reported in the Performance report under the Web search type, together with ordinary results. An AI Overview occupies a single position, and a click on an external link inside it counts as a click; the help pages describe no filter that separates them.
Google says there are no additional requirements or special optimisations for AI Overviews: a page must be indexed and eligible to be shown with a snippet. For ChatGPT search, OpenAI recommends allowing its OAI-SearchBot crawler in your robots.txt file.
Monthly is enough for most small and mid-size websites: citations and AI Overviews on a fixed set of searches, and AI-assistant visits by landing page. Check crawler access again whenever robots.txt or the site's templates change.
Run the routine by hand for two months to learn which signal matters for your site. To have the citation checks, AI-assistant visits and AI Overview notes gathered for you, request access to Website Growth Ops through ITMTB. If you would rather ITMTB set up the measurement and work through the fixes with your team, see our analytics and AI services or contact us.
ITMTB works with enterprises on AI, cloud, security, and digital transformation — from strategy through deployment.