How AI Overviews Changed Search Marketing

What changed, stated plainly

The economics of a search result changed. The mechanics of earning one did not.

A page still has to be crawled, rendered, indexed and judged more deserving than the alternatives, and the documented requirements for appearing in AI features are the ordinary requirements for appearing in search at all. What changed is what a position is worth. A result page that answers the question in place converts fewer of its impressions into visits, which means the same ranking produces less traffic than it did, and targets set on the old conversion rate will be missed by teams doing nothing wrong.

Almost everything sold as a response to this conflates those two things — treating a change in click economics as evidence that the technical work has changed. It has not, and the confusion is expensive, because it redirects budget from things that work toward products built on an assumption nobody has tested.

The zero-click figure, and the two things that must travel with it

The number in circulation is this: 68.01% of US Google searches ended without a click between January and April 2026, leaving 31.99% that produced at least one click. It comes from SparkToro's analysis of Similarweb clickstream data, and it is the most widely quoted statistic in search marketing right now. It is also the most widely misquoted, because two caveats are almost always stripped off.

First, the panel. In the source's own words, the data "is not inclusive of the Google mobile search application, but rather mobile searches within the browser." It is a desktop and mobile-web panel. SparkToro adds that because zero-click features are used even more aggressively inside the search app, the true figure is likely higher — but that is their inference, not a measurement.

Second, the trend is not like-for-like. The multi-year comparison combines data from different panel providers across different years, and SparkToro says so directly: "these aren't the same users or devices, they're not even necessarily perfect demographic matches." Which means the honest use of this figure is as a snapshot of a period, never as a measured change from an earlier one. Anyone presenting a year-over-year delta from these panels is presenting an artifact of the methodology as a finding. The study states both caveats itself.

What an AI Overview actually costs in clicks

For the narrower question — what happens to a top-ranking page when an AI Overview appears above it — the best-evidenced figure comes from a study of 300,000 keywords, half with an AI Overview present and half without, comparing aggregated Search Console desktop click-through rates for March 2024 against March 2025.

The finding: informational keywords showing an AI Overview had a 34.5% lower average click-through rate for the top-ranking page than comparable informational keywords without one. Underneath that headline, position-one click-through for informational keywords fell from 0.056 to 0.031 across the two periods, and for the AI Overview subset specifically from 0.073 to 0.026.

The authors' own qualifications matter as much as the number. Site-level reports vary widely, roughly 20% to 40%. Multiple citations inside a single AI Overview dilute clicks to any one of them. And Google provides no way to isolate AI Overview performance in Search Console — still true for clicks, even after the June 2026 reporting change. Treat 34.5% as the best available estimate for informational queries in that window, not as a constant to apply to your own forecast. Larger figures circulate; the ones I have seen do not trace back to a published methodology, so they do not appear here. The methodology is published in full.

Prevalence is more volatile than the panic suggests

The framing that AI Overviews are everywhere and growing without limit does not survive contact with the largest published prevalence study, which tracked more than ten million keywords through 2025. Trigger rates ran 6.49% in January 2025, peaked at 24.61% in July, and fell to 15.69% by November. They went down.

The composition shifted at the same time, and this part matters more for planning than the headline rate. Informational queries fell from 91.3% of AI Overview triggers to 57.1%, while commercial rose from 8.15% to 18.57%, transactional from 1.98% to 13.94% and navigational from 0.84% to 10.33%. The feature moved toward the queries with commercial intent — the ones marketers care most about — even as its overall footprint shrank. Triggering keywords also skew long-tail: nearly 60% attract a hundred or fewer monthly searches.

And one finding worth carrying precisely because it is inconvenient for the alarming version of this story: tracking the same keywords before and after an AI Overview appeared, that study found zero-click rate decreased slightly, from 33.75% to 31.53%. That does not contradict the 68.01% figure — the two measure different things, one the share of queries showing the feature and the other the share of searches producing no click — but it does mean the simple narrative is not supported by the largest dataset on prevalence. Both numbers belong in the same report.

Why the headline percentage is not your percentage

Prevalence is category-specific and it moves, which makes any published rate a poor substitute for measuring your own query set. On the keyword set researched for this site, 870 of 1,264 queries returned an AI Overview — 69% — well above the published average for that period, because this subject area is unusually informational. A transactional catalog in a regulated industry could sit far below it.

So the first analytical step is not adopting a figure, it is producing one. Take the queries that actually matter commercially, record whether each returns an AI Overview, date the record, and repeat it. That gives you an exposure profile you can plan against, and it separates the queries where a click is still likely from the ones where the answer is now given in place. Those two groups deserve different content, different targets and, frequently, a different channel.

What this changes about targets and reporting

Four practical adjustments, all of which are about honesty rather than technique.

Stop treating position as the outcome. It was always a proxy; it is now a weaker one. Impressions tell you whether the site is being considered, clicks tell you whether it is being chosen, and the ratio between them is where AI features show up in your data.

Expect click-through to decline on informational queries without a ranking change. If impressions hold, position holds, and clicks fall on informational terms, that is the visible signature of the result page changing around you rather than a failure of the work.

Report AI appearance separately, and only as appearance. Search Console's generative AI report gives impressions, pages, countries, devices and dates, with AI Overviews and AI Mode combined and no click data. Reporting it alongside traffic without that caveat invites a conclusion the data cannot support.

Weight measurement toward what happens after the click. Fewer visits of higher intent is a plausible outcome of all this, and a report built on sessions will read it as a loss while a report built on leads or revenue will read it correctly.

What did not change, which is most of it

The technical work is untouched. Pages still have to be crawlable, renderable and indexable; the documented prerequisite for appearing in AI features is being indexed and eligible to be shown with a snippet, which is the same eligibility that governs ordinary results. Google states that there are no additional requirements and no special optimizations, and that no new files, AI text files or schema.org markup are needed.

Confirmed core and spam updates remain the first thing to rule out when traffic changes, and they are routinely misattributed to AI features by people working without dated data. Google publishes a dated history of every confirmed update, and reading it before diagnosing is the difference between an answer and a guess. Spam policies still apply, including scaled content abuse — where the qualifier is generating many pages without adding value, not the use of AI as such — and site reputation abuse.

It is also worth being careful about a related claim. The 2025 antitrust remedies, which ban exclusive default search contracts and require Google to share portions of its index and interaction data with qualified competitors, have no measurable effect on how pages are ranked. Recommendations attributed to that ruling are commentary, not consequence. The plausible medium-term effect is more viable non-Google surfaces receiving that data, which would make diversification less speculative than it is today — but that is a forecast, and should be labeled as one.

How to tell whether AI Overviews are actually costing you

The honest answer is that you cannot isolate it precisely, because Google does not report clicks from AI features and does not let you segment results by whether an AI Overview appeared. What you can do is triangulate, and it is worth doing before you act on anyone's assertion.

Segment your queries by intent, since the measured impact is concentrated on informational terms. Within that segment, look for the pattern of stable impressions and stable position with falling click-through over a window long enough to exclude seasonality. Cross-reference the dates against the published history of ranking updates, because a core or spam update inside the window is the more common explanation and the two are routinely confused. Record whether an AI Overview currently appears for those queries, with a date, and repeat the check, since prevalence moves.

If that pattern holds — informational queries, position stable, click-through falling, no update in the window — you have reasonable evidence. It is still evidence rather than measurement, and saying so is what separates a useful report from a confident one.

Frequently Asked Questions

Published 2026-08-11