How to Measure AI Search Visibility
ChatGPT cited the exact same websites on consecutive days only 2% of the time.
That’s one of the findings from our recent State of AI Search report. If you’re trying to work out how to measure AI visibility by typing prompts into ChatGPT, that number should worry you.
Without reliable measurement, you can’t tell whether your marketing changes are influencing your AI Search performance. As Tim Cameron-Kitchen, Founder of Exposure Ninja, puts it, without measurement, “we’re essentially flying blind.”
Here’s how to measure AI Search visibility and close the gap between what AI platforms recommend and what your analytics actually shows.
Why Manual Prompt Checks Can’t Measure AI Visibility
The old approach was simple: type a question into ChatGPT and see if your brand comes up.
Tim Cameron-Kitchen explains why that no longer works: “Measurement has moved beyond the days where you’d type something into ChatGPT and just see if your brand appears. This is an incredibly flawed way to do it because these AI answers are created fresh each time, and they change a lot.”
The data from our State of AI Search report backs this up. Google AI Overviews cited the same websites on consecutive days just 6% of the time. When the same prompt was run on ChatGPT and AI Overviews, 49% of the time the two platforms had no websites in common.
In other words, “most of the time, these AI platforms are citing different websites each day, even for the same prompt.”
The fix is volume. Patterns only appear when you collect enough data across enough prompts, which is why marketing teams need a dedicated AI visibility tool rather than a spreadsheet of manual checks.
Set a Baseline With Share of Voice, Mentions and Citations
Tools like Ahrefs Brand Radar* let you track your visibility across multiple AI platforms and watch how it changes over time, for your brand and your competitors.
Tim Cameron-Kitchen explains why that baseline matters: “Anything we do in marketing, we want to baseline our current performance so that we can then fix the gaps and see how those fixes change over time.”
In the video demo, Tim Cameron-Kitchen runs Brand Radar on Together Money, a UK specialist lender. The tool’s suggested competitors were too large and too broad for a fair comparison, so he swapped in direct rivals Pepper Money and West One Loans.
Location matters just as much. Compare worldwide against a competitor operating in several countries and “they’re obviously going to absolutely smash you.”
Once set up, you can track three core metrics:
- Mentions: How often your brand is recommended in AI answers.
- Citations: How often your website is used as a source.
- Impressions: How often your brand is seen in an AI answer.
You can also break these down by platform, which helps if you know where your buyers spend their time.
“Maybe they’re all on Microsoft, sat at their desks in the office, so therefore there’s going to be a lot more Copilot searches, or maybe you know that your audience is B2C, and they’re going to be using ChatGPT. You might want to prioritise tracking your visibility and performance in some AI platforms over others,” explains Tim Cameron-Kitchen.
Filter for the Prompts Most Likely to Drive Revenue
The AI Responses data can be overwhelming. In the Together Money example, there were tens of thousands of pages of results.
The way through is to filter by commercial intent and funnel stage.
For Together Money, that surfaced prompts like “who offers second charge mortgages” and “commercial mortgage lenders UK”. These are high-value searches from people close to a purchase decision.
Add the “Not Mentioned” filter, and you get your biggest opportunities: commercial prompts where competitors appear and you don’t. Content targeting these could be “real untapped potential of driving some high-quality commercial decision-focused traffic to our website.”
If individual prompts feel like too much, step up a level to topics. Just be prepared for some odd results. Tim Cameron-Kitchen spotted National Rail, best friend gifts, and spouse visas in a mortgage lender’s competitive set.
“This is when you need to use your judgement as a marketer to understand, do I really want to compete for these topics, or have I just picked up some visibility accidentally, or are my competitors picking up some weird visibility because they’ve published some random content years ago?”
Turn Cited Pages Into Digital PR and Content Fix Lists
Citations matter because they usually come with a link. That link is how AI visibility becomes website traffic.
The cited pages report shows which articles, listings, and videos AI tools draw from in your space. Filter by pages where your brand is not mentioned, and you have a digital PR target list.
“These are the pages that we want our brand to be found on because these are the pages that are influencing the AI answers. And if we can get these pages talking about our brand, we stand a better chance of being recommended by the AI tools,” Tim Cameron-Kitchen, Founder of Exposure Ninja.
Found But Not Cited: Your On-Site Fix List
The “Found but Not Cited” filter shows pages on your site that AI tools found during research but chose not to use. Tim Cameron-Kitchen explains the likely causes:
“This can be because it finds the information confusing, or it finds the information low quality, or maybe we need to put an AI-focused summary at the top, or signpost that page better with headings, or give that page some more value, like first-party data.”
He uses Currys’ “Best Ovens in 2026” page as an example. It exists to attract oven shoppers, and AI tools are finding it. “The bad news is they’re not citing it.” Improving that one page could mean more links in AI answers and more buyers landing on it.
“Doing this as a one-off exercise is great, but it’s even more valuable to track these things over time, because that allows you to see how the work that you’re doing on the website and getting cited in these other websites is influencing your performance,” Tim Cameron-Kitchen, Founder, Exposure Ninja.
Why GA4 Only Shows a Fraction of Your AI Search Traffic
Google Analytics 4 (GA4) now reports AI assistant traffic separately. In one example Tim Cameron-Kitchen shares, just 136 sessions came from AI tools, but over 20% of that traffic converted.
That’s impressive, but it’s only a small slice. The rest is dark traffic.
“When you’re recommended a brand in ChatGPT, you don’t always click a link straight to that brand’s website in order to buy. You might forget about it completely and then come back to it later on and do a search, or type the brand into your browser, or go and follow them on social media.”
Similarweb’s research on the downstream impact of AI visibility found that after an AI recommendation, people searched for the brand 56% of the time and went directly to its website 20% of the time. Only 8.8% visited via an AI tool.
On top of that, Google AI Overviews are currently lumped in with organic search, not AI assistant traffic. Rely on GA4 alone and “we’re missing the majority of the impact of AI Search.”
Tim Cameron-Kitchen illustrates this with Elite Renewables, a brand Exposure Ninja owns. A user asks ChatGPT about replacing a gas boiler, gets a shortlist of heat pump installers including Elite, then doesn’t click.
“Later on, they just go to Google and search Elite Renewables, or they have a look at one of Elite’s videos on YouTube, and then they become a lead later on. The Elite marketing team would have no idea that that search has been influenced by this AI visibility, and that’s the measurement gap,” Tim Cameron-Kitchen, Founder, Exposure Ninja.
AI Search Attribution Is Improving, but Don’t Wait for It
Exposure Ninja expects the gap to start closing, though “we’re not saying that perfect attribution is around the corner.”
The pressure is coming from regulators. On 3 June 2026, the UK’s Competition and Markets Authority (CMA) demanded Google give publishers more data on their AI visibility, including click-through data separated from general search. The same day, Google rolled out generative AI performance reports in Search Console.
Those reports currently show impressions, not clicks. Our State of AI Search report predicts platforms and analytics providers will improve how AI visibility and referral traffic are reported, with click data a likely next step in 2027.
Tim Cameron-Kitchen’s advice is clear: “Don’t wait for perfect attribution. We can already see that user behaviour is going towards AI Search, and the early indications are that this traffic and this visibility are super, super valuable.”
How to Fill the Attribution Gap Today
Track share of voice, mentions and citations. Monitor AI assistant traffic in GA4, remembering much of it will show up as direct and organic. If you have a sales team, ask every new lead where they found you.
At Elite Renewables, “we consistently hear ChatGPT being the top source of people finding us. But if you look in analytics, it’s down at about source number seven.”
If improving your visibility in AI Search is a priority for the year ahead, our expert team can help. Book a consultation call and tell us about your goals and current challenges. One of our consultants will identify your top AI Search Optimisation opportunities to get recommended in tools like ChatGPT, and turn that visibility into qualified leads.
*Ahrefs is the sponsor of this video. Links to Ahrefs Brand Radar may be sponsored or affiliate links: confirm disclosure treatment before publishing.
