Position one on desktop earned 7.43% of clicks in August 2026. Twelve months earlier it earned 17.83%. That is the headline from DataForSEO’s Search Insider: 2026 Edition, built from organic performance data across more than 2,000 websites between September 2025 and August 2026.
The number everyone will quote is the 10.41 percentage point drop. The more useful number sits next to it. Over the same year, mobile position one fell from 25.17% to 20.80%, a drop of 4.37 points. Mobile position two actually went up, by 0.53 points.
So the click collapse is mostly a desktop event. Almost nobody is reporting it that way, because almost every CTR study publishes one blended figure.
The full picture, both devices
Desktop, September 2025 to August 2026:
| Position | Sep 2025 | Aug 2026 | Change |
|---|---|---|---|
| 1 | 17.83% | 7.43% | −10.41pp |
| 2 | 15.27% | 3.07% | −12.20pp |
| 3 | 3.43% | 0.58% | −2.85pp |
| 4 to 10 | 1.23% | 0.60% | −0.63pp |
Mobile, same window:
| Position | Sep 2025 | Aug 2026 | Change |
|---|---|---|---|
| 1 | 25.17% | 20.80% | −4.37pp |
| 2 | 7.99% | 8.52% | +0.53pp |
| 3 | 5.98% | 4.77% | −1.21pp |
| 4 to 10 | 3.08% | 1.93% | −1.15pp |
Position two on desktop lost more in absolute terms than position one. It went from 15.27% to 3.07%, which leaves it earning roughly a fifth of what it earned a year ago.
It happened in March, not over a year
The year-on-year figure makes this look like erosion. The monthly table tells a different story.
Desktop position one, month by month: 17.83, 16.54, 18.24, 15.34, 13.33, 15.18. Noisy, drifting down, nothing dramatic. Then March 2026 comes in at 7.94%, and the rest of the year reads 7.21, 8.63, 9.27, 8.91, 7.43.
That is a cliff in a single month followed by eleven months at the new level. Desktop position two had its own cliff earlier, dropping from 15.27% in September 2025 to 5.37% in October 2025.
Mobile did something different again. Position one fell to 12.96% in February 2026, then climbed back to 22.49% by May and finished the year at 20.80%. It dipped and recovered. Desktop dropped and stayed down.
I would be careful about naming a single cause. A panel of 2,000 sites can shift because the mix of sites in it shifted, and DataForSEO do not claim to have isolated a cause. What I will say is that a step change in one month looks far more like a layout or measurement change than like a million people slowly changing their habits.
What the AI Overview actually takes
The same report estimates where clicks go when an AI Overview is present, with zero-click searches excluded.
AI Overview citations take 42% to 45% of the clicks. The top organic result drops from a 33% to 38% share down to 21% to 24%. Positions 2 to 10 lose roughly 36% to 42% of their combined share.
The consistency is the interesting bit. Whether someone is researching, comparing, navigating or buying, the AI Overview takes about the same 42% to 45%. You would expect transactional queries to hold up better, because a buyer with a card in their hand wants a website rather than a summary. On this data they don’t.
How it squares with everyone else’s numbers
No single study should carry an argument this big, so here is where the others landed.
Ahrefs re-ran their analysis in February 2026 and found an AI Overview correlates with a 58% lower average click-through rate for the top-ranking page, up from 34.5% in their April 2025 run. Seer Interactive, tracking 5.47 million queries, reported organic CTR on AI Overview queries falling from 1.76% to 0.61%. Pew Research’s panel work found users clicked an organic result about 8% of the time when an AI Overview appeared against 15% without. Sistrix, looking at Germany, put position one at roughly 27% falling to 11% with an AI Overview present.
Four different methods, four different numbers, one direction. Most of those figures reach me through secondary write-ups rather than the original reports, so I would treat them as corroboration of the shape rather than as precise values.
One thing I would flag about the DataForSEO desktop curve, because nobody else seems to have: position two starting at 15.27% against position one at 17.83% is a very flat top of funnel, and position three at 3.43% is a very steep fall straight after. Real CTR curves rarely look like that. It does not invalidate the year-on-year change, which is measured consistently against the same baseline, but it is a reason to use the direction and not the decimal places.
A caveat that matters more than it sounds
This is United States, English-language data. Only the backlink portion of the report is global.
We sell to UK businesses, so I am not going to put an American CTR curve into a British forecast and call it research. AI Overview rollout timing, competitor density and the share of search that runs through Bing all differ here. The UK picture is thinner because almost nobody publishes it, which is part of why Bing is half our search traffic came as a surprise to people when we wrote it up.
Use your own Search Console. It is the only CTR curve with your name on it.
What to do with this on Monday morning
Split your Search Console CTR by device before you do anything else. If your buyers research on desktop at work, which most B2B and regulated-sector buyers do, you have been reading a blended number that flatters you.
Stop converting rankings into traffic forecasts. The multiplier that turned position one and a search volume into an expected click count is the thing that broke. Rankings still matter. The arithmetic behind them doesn’t hold.
Move your reporting to impressions and position. Both survive the click collapse. Clicks are now the least stable number on the page.
Then work out whether you are inside the answer. If an AI Overview takes 42% to 45% of the clicks, the question worth answering is whether you are one of the sources it was built from. That is a different discipline, and we set out how we approach it in the Synaptic Authority Engine.
If you want to see where you currently stand across the AI platforms rather than guess, start with the AI Visibility Score. It takes a few minutes and it costs nothing.