Are AI overviews taking our clicks?

This is an AI-search strategy question — the kind that usually shows up from leadership, content, client. It rarely has a one-line answer, because the honest version of “Are AI overviews taking our clicks” is a shortlist of rival explanations, not a single cause. The job is to work through that shortlist with evidence and stop as soon as one of them is confirmed — not to write a report that mentions all of them.

The rival explanations

Direct answer: Five distinct patterns explain what an AI Overview does to a given query — informational-query click loss, citation-without-click brand presence, impression inflation masking a CTR collapse, original-data pages getting cited more than recaps, and local/product queries behaving differently than national informational ones — and treating all queries the same way misreads which pattern is actually in play.

Treat these as competitors, not a checklist. The point of naming five up front is to stop the first plausible-sounding one from becoming the story before the others have been checked.

• Informational queries lose clicks; commercial investigations still click.
• We are cited but not clicked — brand is present, traffic is not.
• Impression inflation hides a CTR collapse on the same average position.
• Original data and first-party tools get cited more than recaps.
• Local / product queries behave unlike national informational ones.

What the evidence has to show

Direct answer: A frozen weekly query set tracking impressions-up/clicks-down, a manual or scripted Overview-presence-and-citation sample, cross-engine citation checks in Gemini/ChatGPT/Perplexity, landing-page engagement after Overview-prone queries, and treating vendor AI-visibility scores as directional-only are what classify a query correctly.

None of the five above survives on a hunch. Here is what actually needs pulling before any of them can be ruled in or out:

• GSC: impressions up / clicks down on a frozen query set, weekly.
• Manual or scripted sample: Overview present Y/N, our citation Y/N, our classic rank.
• Citation checks for brand + SMEs in Gemini, ChatGPT, Perplexity on 25 prompts.
• Landing-page engagement after Overview-prone queries.
• Vendor AI-visibility scores treated as directional only.

The decision rule

Direct answer: Classify every money query as classic-SERP, Overview-prone, or mixed. Optimise Overview-prone queries for citation and brand search, not for yesterday’s CTR targets. Do not panic-rewrite pages whose classic rank is stable and commercial.

What to tell the people around you

Direct answer: Leadership needs a query-class dashboard and funding for citable original data — not a panic rewrite of pages whose classic ranking and commercial performance are actually stable.

The analysis is not finished until it produces something a non-specialist can act on. That means naming the situation, the cost of getting the first move wrong, and a specific ask — not a summary of the investigation.

• Situation — Generative answers are changing click economics on some query classes, not all of search.
• So what — Using 2022 CTR curves to judge 2026 pages will fire the wrong teams. Class-level measurement is the fix.
• The ask — Accept a query-class dashboard. Fund original data that can be cited. Stop judging informational URLs on click-through alone.

SEO measurement. Content adjusts formats by class.

How to act on this

1. Classify every money query as classic-SERP, Overview-prone, or mixed using a frozen weekly tracking set.
2. Sample Overview presence and your own citation status manually or via script, alongside your classic ranking on the same query.
3. Check citation status for the brand and named subject-matter experts across the major AI engines, treating any vendor score as directional only.
4. Review landing-page engagement specifically for sessions that survive an Overview-prone query, not just overall traffic.
5. Fund original data or first-party research on the queries most likely to be cited, and stop judging purely informational pages by click-through rate alone.

Frequently asked questions

Does an AI Overview always reduce clicks for a query?

No — commercial and investigative queries tend to still generate clicks even with an Overview present, while purely informational queries lose more. Classification by query type is what separates the two.

If we're cited in an Overview but not clicked, is that worthless?

It's a different kind of value — brand visibility and citation exposure without a click. It shouldn't be reported as a traffic win, but it also shouldn't be reported as a total loss; track it as its own signal.

Should we rewrite every page that's losing clicks to an Overview?

Not automatically — first confirm the classic ranking and commercial performance are actually degrading. A stable-ranking, stable-converting page experiencing an Overview-driven impression/click shift doesn't need a content rewrite.

How reliable are third-party AI-visibility scoring tools?

Treat them as directional signals only, not as a precise measurement — the methodology behind most of these scores isn't independently verifiable, so weight them alongside, not instead of, your own citation sampling.

Do local and product queries behave the same way as informational ones under AI Overviews?

No — they tend to follow different click patterns, which is why the classification step separates query classes rather than applying one universal rule across the whole query set.

Will this classification stay stable over time?

No — Google continues to expand where Overviews appear, so the frozen query set needs periodic re-classification rather than a one-time labeling exercise.

Should we try to prevent our content from being used in AI Overviews?

Blocking AI crawlers usually means losing citation opportunities entirely, not preserving clicks — the more productive response is optimizing for citation and brand visibility within Overview-prone queries rather than opting out.

Sources

1. How Google's ranking systems work — Search Central
2. Creating helpful, reliable, people-first content
3. Search engine results page — Wikipedia
4. Generative artificial intelligence — Wikipedia

Where nqzai fits

nqzai runs this same rival-hypothesis framework against your own connected Search Console, Analytics, and audit history, and returns a keep / change / stop decision with the evidence named — including which of the explanations above it could not test, and what to connect to close that gap. No extra cost for the analysis itself; it reads measurements already on file.

Ask nqzai: “Are AI overviews taking our clicks?”