AI Overviews Is Lying to Your Analytics. Straight to Your Face
If you're sitting in Search Console waiting for Google to honestly show you traffic from AI Overview — stop waiting. It won't show you. Ever.
A study just came out that tracked AI Overview data for nine months straight — over 51,000 tracked events. And there's one number in it that should worry you: 22% of AI Overview traffic gets attributed in analytics as Direct, not Organic. Almost a quarter of your clicks from AI-generated results are disguising themselves as people typing your URL straight into the browser. Your report is lying to your face, and you're sitting there celebrating "direct traffic" growth.
Why this happens
The mechanics are simple and cynical at the same time. When a user clicks a link inside AI Overview, the transition often happens without a proper referrer — the browser or Google's own interface doesn't pass the source the way a normal organic click does. Analytics sees an empty referrer and logs the visit as Direct. Technically, it's not lying. Practically, it's distorting the whole picture.
And Google gives you zero native signal for AI Overview in Search Console. No separate report, no filter, not even a hint. This isn't a bug — it's a position. The less you see, the fewer questions you ask about how your content actually gets cited inside AI answers.
What AI Overview actually cites
The same nine-month dataset shows a clear pattern: structured content gets cited most — comparison tables, prices, delivery timelines, clear specs. Not vague "everything about everything" overview articles, but concrete answers to concrete questions. The logic makes sense: it's much easier for a model to pull a ready fact out of an HTML table than to dig through five paragraphs of fluff to find a number.
One more thing that matters — AI Overview traffic is unstable. A citation is there today, gone tomorrow, back next week in a different wording. This isn't a "write one article, get cited forever" story. It's a living process with seasonality and a lifecycle, something you need to monitor constantly, not set up once and forget.
What to actually do about it
Since Google won't give you the signal, you build your own. Here's a practical setup: create a custom dimension in GA4 that catches URLs containing the #:~:text= fragment — that's exactly how links get tagged when someone clicks through from AI Overview or similar AI-generated snippets. It's not a perfect fix, but it's a real fix — one that gives you an actual picture instead of a pretty lie about "Direct."
Next — content. If you run e-commerce, a travel or transportation site, a GEO project — turn your key comparisons, prices and timelines into HTML tables. Not images, not PDFs — actual marked-up HTML that's easy to parse. This isn't about looking nice. It's about making sure the model can pull your fact instead of your competitor's.
And last — stop writing vague "about everything and nothing" articles. AI Overview likes clear answers to clearly asked questions. This, by the way, is an old story for anyone who's been in SEO long enough: precision has always beaten fluff. It used to just affect rankings. Now it also decides whether you get cited at all.
Prioritize content based on data, not gut feeling — on what's actually getting cited and how often. At least once a quarter, check which pages are losing citations and which are gaining them, and update whatever's dropping first.
My take
Nobody owes you transparent analytics for free. Google certainly won't hand it to you — it benefits from you not understanding where your traffic actually comes from. So you count it yourself. Yes, it's extra work. But the alternative is running your site blind, trusting a report that inflates Direct and stays silent about AI Overview. If you want to sort this out properly on your own project, that's exactly what our SEO audit and analytics setup services are for.