Your "AI Visibility" Score Is a Guess: How to Report AI Search Numbers Honestly
Analytics
Step-by-step
AI visibility

Trigger Marketing
The Trigger team
⊹
Somebody somewhere is putting "AI visibility is WAY up!" on a slide right now. With a rocket emoji.
Here's the problem. Nobody publishes how often people ask ChatGPT anything. So every "AI search volume" and "AI visibility" number in an SEO tool is built by a model. It's a guess wearing a suit.
That's the main point from SEO consultant André Pitì on the Candour podcast (Nov 2, 2026). Use the numbers. Just never make them the headline. Here's how.
Quick words (one line each):
AI Overview: the AI answer Google shows at the top of some search results.
GA4: Google Analytics 4, the free tool that counts visits to your site.
Server logs: your web server's diary of every visit, including bots.
UTM tag: the "?utm_source=..." bit added to a link so analytics knows where a click came from.
Channel: how GA4 groups traffic (search, social, referral, direct and so on).
Step 1: Know how the sausage gets made
André Pitì breaks down two common tool numbers:
"AI Overview visibility" is a snapshot. The tool checks a keyword once, sees if you were cited, and credits you with that keyword's Google volume. That changes by country, device and login. For low-volume business keywords, he found data from 2023-2024, before AI Overviews even existed.
"AI search volume" is Google volume times one multiplier. Tools take a related keyword's Google volume and shrink it by how big the AI platform is next to Google. One multiplier for every topic treats a compliance question and a recipe question the same.
Step 2: Label every tool number "estimate"
In every report, write "tool estimate" next to AI visibility or AI volume.
Put your own Search Console and GA4 numbers right beside it.
Report the direction ("going up", "flat"), not one exact figure.
Step 3: Don't call a "direct" jump AI traffic
A jump in "direct" visits proves nothing by itself. App clicks, privacy settings and people typing your address all land there. Instead:
Look at which pages got the direct visits.
Nobody types a long /blog/... address by hand. Deep pages getting "direct" visits is a clue.
If those same pages are the ones AI cites, now you have a stronger clue.
Step 4: Make sure your site keeps UTM tags
If your site redirects any link with "?utm_..." to the clean address, the tag gets stripped before GA4 sees it.
Open a page with a tag, like yoursite.com/?utm_source=chatgpt.com
Test it in Search Console's live URL test.
You want a 200 (page loads), not a 301 or 302 (redirect).
Keep the canonical tag pointing to the clean address. Just don't redirect.
Step 5: Cross-check server logs with GA4
If AI user bots (like ChatGPT-User or Perplexity-User) spike on certain pages, and direct visits rise on those same pages, you can make a careful guess. Two clues beat one.
Skip bot locations in your logs, though. AI bots come from data centers, so their location tells you nothing about real people.
Step 6: Expect your AI numbers to disagree
Host Jack shared his own site's numbers (Nov 2, 2026): a regex-based AI channel in GA4 showed about 500 visits, while GA4's built-in AI assistant channel showed about 2,000. Same site. Don't trust just one.
While you're in GA4:
Move your AI channel above "Referral" in the channel order.
Add Country as a second column.
Step 7: Check where AI impressions come from
Open Search Console's generative AI report.
Look at impressions by country.
André Pitì's client looked great, until they found a same-name company in Chile was getting the impressions.
Step 8: Mine long, prompt-like searches
In Search Console, filter for queries of 8-9 words or more. Some run 70+ words.
Group them by ideal customer. Toss the off-topic ones.
Split by country before you decide which pages to improve.
Most will have zero clicks. They're a clue about what people ask, not proof of traffic.
Step 9: Ask better questions
André Pitì's new questions: How often and how consistently are we mentioned? Are the facts about us correct and up to date? Old affiliate links, old listicles and bad reviews on G2, Capterra or Clutch can feed AI wrong facts.
Do this today
Write "tool estimate" next to every AI number in your report.
Test one page with a UTM tag in Search Console: you want a 200.
In GA4, move your AI channel above Referral.
Check AI impressions by country.
Report the trend, not the exact number.
Related reading
Stop screenshotting ChatGPT: how to measure AI search traffic
Your rank tracker says you tanked? Check Search Console first
Sources
Don't worry if this seems like it's a pain in the ass for someone who doesn't build websites. That's because it is. The good news: just click the link below, pull the trigger, and we'll take care of the rest.
Share This Article

