Google AI Overviews and your traffic: measure the drop before you panic

Google's AI Overviews answer many searches directly on the results page. For publishers that means fewer clicks on the same rankings — the traffic-panic story of the past two years. What's usually missing from that story: the effect varies wildly by site and by query, and you can measure yours instead of borrowing someone else's headline.

The signature: impressions hold, clicks fall

An AI Overview doesn't remove you from results — it answers above them. So the fingerprint in Search Console is specific:

1. Open Search Console → Performance, compare the last 3 months to the prior 3.
2. Sort queries by click loss, then check each one's impressions.
3. The AI-Overview signature: impressions flat or up, clicks and CTR down, position unchanged. You still rank; the click just got answered away.

Queries with falling impressions or position are a different problem (ranking loss — ordinary SEO). Keep the two piles separate or you'll fix the wrong thing.

Which queries lose, which don't

The pattern across affected sites is consistent with what the Overview format can do:

• Definitional and how-to queries ("what is X", "how to do Y") — hit hardest. The answer fits in a paragraph, so the paragraph appears on the results page.
• Transactional and navigational queries — barely affected. Google still wants the click to happen for "buy", "pricing", "login", brand names.
• Complex, multi-factor decisions — partially affected; Overviews summarize but people still click through for depth, tools, and trust.

That mapping is your triage list: content whose entire value fits in a paragraph is structurally exposed. Content with depth, tools, data, or a product behind it keeps its clicks.

The part analytics can't see

Being cited by an AI Overview — or read by Google's AI fetchers when Gemini answers about you — produces visits and mentions your analytics either buries in Direct or never sees at all. Before concluding "AI killed our traffic," check what AI traffic you actually receive: crawlers, assistant fetches, and AI referrals hiding in Direct. Several site owners discover the AI channel is sending them qualified visitors at the same time Overviews shave commodity clicks. You can't make that trade-off consciously if you only measure one side.

What to actually do

Reshape exposed content around what a paragraph can't deliver. Add the calculator, the downloadable, the data table, the tool. AI Overviews summarize; they don't execute. (It's the strategy behind our own free tools — a summary can't check your logs for you.)

Feed the machines the facts you want repeated. Clean HTML, accurate structured data, an llms.txt, and crawlable canonical fact pages make the difference between being paraphrased accurately and being paraphrased wrong.

Keep serving the clicks you still get. Falling CTR makes each remaining click more valuable — landing-page quality and conversion suddenly matter more than raw traffic reporting.

Don't block Google's AI crawlers in retaliation. Blocking Google-Extended only opts you out of model training; it does not remove AI Overviews for queries where you rank, and blocking core Googlebot to escape Overviews de-indexes you entirely. The lever everyone reaches for first is connected to nothing.

The honest framing: AI Overviews tax content whose value was always thin, and they reward sites that know exactly what the AI layer is doing on their pages. Measure first — the data is sitting in Search Console and your own traffic, and it's probably less apocalyptic and more specific than the headlines suggested.