What AI Overviews Did to the Click Curve

When an AI Overview sits above the search results, fewer people click the links beneath it — on the queries where it appears. How much fewer depends on what you measure. Published studies report a lower click-through rate on affected keywords, a smaller share of searches that end in any click, and a click rate that dropped and later steadied. They differ because they measured different things. This piece lays the three side by side, and reads them honestly.

That is the whole mechanism, and it is why the effect is uneven. On a search with no Overview, the position curve still governs, and a page-one ranking is worth what it always was — priced in What Page One Is Worth. On a search where an Overview appears, a narrow, measurable question opens: how much of the click does it take? Several groups have measured it, and their answers only look contradictory until you read what each one counted.

Each row is read from that study’s own published, dated report and stated only for what the study measured. The three are not blended into a single figure.

The newest data adds a wrinkle the alarming headlines tend to drop. In the longest-running series here, the click rate on AI-Overview results fell through 2025 to about 1.3% in December, and by February 2026 had climbed back toward 2.4%. Its own authors call that a leveling off, not a recovery, and caution against forecasting from two months of data. Over the same stretch, the click rate on results with no Overview rose. That is not a collapse to zero; it is a curve finding a new level — a very different story from the one that says search clicks are ending.

1. Which Queries?

   Every search, or only the ones an AI Overview actually appears on? Branded or non-branded? A number for “all searches” and a number for “affected keywords only” are not the same claim, and the second is almost always the larger drop.

2. Which Baseline?

   Measured against what — the same keywords before Overviews existed, a different group of keywords without one, or a forecast of what might have happened? The baseline chosen decides how big the drop looks.

3. Which Period?

   One month or a full year? A single snapshot or a trend? A figure from a two-month window can move a great deal, and a decline measured at one moment can read very differently a quarter later.

4. Cited, or Merely Present?

   A page named inside the answer and a page merely sitting in the results below it earn very different click rates. A blended average buries that divide, which is often where the real story lives.

A percentage quoted with none of these attached — a lone figure that travels from article to article, detached from the study that produced it — tells you almost nothing. In this subject the most-repeated numbers are often the ones whose method no one can name. A figure you cannot trace is not evidence; it is a rumour with a decimal point.

Two consequences follow, both measured rather than predicted. One: on the affected queries, being the page cited inside the answer matters more, because a named source keeps a share of a smaller pool — the same divide the studies above already show. Whether an assistant can reach, read, and name your site at all is the visibility question, walked in Is Your Website Visible to AI? and diagnosed layer by layer in Why Doesn’t ChatGPT Recommend My Business? The other: on the many searches where no Overview appears, nothing here changes — the position curve still rules, and page one is worth what it was.

Read it as a signal, not a verdict. Search Console does not separate AI-Overview impressions from the rest of your search data, so the gap is directional, not a clean measurement — the same honest limit behind every figure on this page. The free scan reads the mechanical side of your visibility the same way: what it can measure, it measures; where a thing cannot be known, it says so.

• Amsive — April 2025. Google AI Overviews: New CTR Research. Average organic CTR across 700,000 keywords, split branded versus non-branded. A 15.49% average fall where an AI Overview appears, 19.98% on non-branded keywords, against an 18.68% rise on the branded keywords that rarely trigger one.
• Pew Research Center — July 2025. Google Users Are Less Likely to Click on Links When an AI Summary Appears. A real browsing panel of 900 U.S. adults over 68,879 searches in March 2025. A search result was clicked in 8% of visits with an AI summary present versus 15% without, and a link inside the summary in 1%.
• Seer Interactive — April 2026. AIO Impact on Google CTR: 2026 Update. Organic CTR per impression across 53 brands and 2.43 billion impressions, January 2025 to February 2026. Click rate on AI-Overview results near 1.3% in December 2025 and about 2.4% by February 2026, with cited brands earning roughly 120% more organic clicks per impression than uncited ones.
• The baseline curve — the position-one click-through figure used above is the same published click-through research cited in What Page One Is Worth, read from the same sourced-data library so the two pages can never disagree.
• On method — each figure is read from the named study’s own report, dated, and stated only for what that study measured. The three are not merged into one number, because they measured different things: a keyword click-through rate, a share of search visits, and a click rate over time. Where a study does not publish a detail, we do not fill the gap with a guess.
• On what cannot be isolated — no public tool yet separates AI-Overview clicks and impressions from the rest of a site’s search data, which is why the self-test above reads as a directional signal and is labelled that way.

Published 3 September 2026. These studies are dated; where a newer reading exists, the date on this page is how we keep the comparison honest.