What Does Reduced Click-Through Mean for Your SEO Strategy?

What did the new click-through study actually find?

Google's AI has are cutting the number of people who ever leave Google. On August 18, 2026, a five-researcher team published a preregistered field experiment on Google Search with 1,100 completed participants, randomly assigned across three conditions over seven days: a stripped-down Search with AI Overviews and AI Mode removed, ordinary Google Search as it ships today, and a version where every query was forced through AI Mode.

The results were stark. Removing AI Overviews raised external click-through rate by 8.8 percentage points compared with normal search. Forcing AI Mode did the opposite: external click-through fell by 18.8 percentage points, daily search sessions dropped, and the sharpest declines hit clicks to news sites, Reddit, and Wikipedia, as Search Engine Journal reported on September 21, 2026.

That is close to a 28-point swing between the two extremes, measured on the same platform, in the same week, against the same query mix. It is the clearest causal evidence yet that AI answers, not just an algorithm update, are behind the traffic a site is losing.

Who does the click-through drop affect most?

Everyone who depends on a click loses something, but the damage is not spread evenly. Publishers absorb the most direct hit: the study found users forced into AI Mode were markedly less likely to reach a news site at all.

B2B SaaS and tech growth teams feel it on comparison and 'best tool' queries, the exact prompts where a synthesized answer can name a winner without sending anyone to a homepage. Local, multi-location, and service businesses feel it differently. An AI Overview can answer 'best plumber in [city]' with a short list and a map, and never route a click to any single business's site.

Older data points the same direction. Ahrefs measured a 34.5% average click-through decline on the top organic result once an AI Overview appears, based on 300,000 U.S. keywords tracked from March 2024 to March 2025. The new experiment adds the missing piece: a controlled test showing the decline is caused by the AI feature, not a coincidence of timing.

Does AI Mode actually improve the search experience?

No. The study's most uncomfortable finding for Google is that AI Mode did not make users happier. People forced into AI Mode reported lower trust in what they found and were more likely to try a competing search engine during the test week.

That matters for strategy, because it removes the easiest excuse for the click-through drop. If AI Mode traded lower publisher traffic for a clearly better user experience, the shift would be a fair exchange. The experiment finds no such trade. Fewer clicks reach the open web, and the people staying on Google are not, by their own report, better served.

What does the before-and-after data show?

Line up the three conditions and the pattern is a straight line, not a cliff. Take the AI has away and clicks to outside sites rise. Add a full AI Mode experience and clicks fall by more than twice as much as they rose when the has were removed.

Google has said publicly that AI Mode improves the search experience. The experiment does not support that framing, and a swing of nearly 28 percentage points between the two arms of the same test is too large for a site relying on organic click-through to treat as background noise.

How should you respond to falling click-through rate?

Stop treating a high ranking as the finish line. A page can hold position one and still lose the click, because the answer already left the results page before anyone scrolled to it. The response is to build for the answer, not just the query.

1. Publish the citable fact first. Give every article one stat, table, or data point with a real, dated source, since that is what a generative engine can quote directly.
2. Answer the question in the first two sentences. Models tend to pull the most direct sentence on the page, not the best-written paragraph three screens down.
3. Track Citation Share alongside click-through rate. A page can be cited across ChatGPT, Perplexity, and AI Overviews without a single visit ever landing in Google Analytics.
4. Cover the full question set in a category, not one article at a time. Coverage and freshness are what get a source chosen and quoted, not a single well-optimized post.

What should you track instead of click-through rate alone?

Click-through rate still matters, but it only measures the fraction of an answer you can see. A fuller picture uses four numbers together.

There is no page two in an AI answer. A brand that is not cited in the answer is invisible to the person asking, regardless of where it ranks on a results page they may never open. Generative Engine Optimization is the discipline built around that fact, and click-through rate is now one signal inside it, not the whole scoreboard.

1. Citation Share: the percentage of relevant AI answers in your category that cite you, the AI-era equivalent of a ranking.
2. Citation Count per day: how often you are cited, which shows volume and momentum.
3. Answer Presence: how much of the underlying question universe in your category you actually cover.
4. Share of Voice: how your citation footprint compares with named competitors on the same prompts.

Sources

Source: arXiv (Wang, Gleason, Bart, Wilson, Metaxa)

Removing AI Overviews raised external click-through by 8.8 points and forced AI Mode cut it by 18.8 points versus normal search arXiv (Wang, Gleason, Bart, Wilson, Metaxa), 2026-08-18

Source: Search Engine Journal

AI Mode did not improve user trust or satisfaction in the study and pushed some users toward competing search engines Search Engine Journal, 2026-09-21

Source: Ahrefs

AI Overviews produce a 34.5% average click-through decline on the top organic result, based on 300,000 U.S. keywords from March 2024 to March 2025 Ahrefs, 2025-04-17