Featured snippets appeared on 0.44% of analysed Google results in August 2026. Two years earlier they appeared on 8.96%. Position zero, the thing a decade of SEO advice told everyone to chase, is effectively over.
That comes from DataForSEO’s Search Insider: 2026 Edition, which tracked SERP feature presence across a monthly panel of 20 million keywords from September 2024 to August 2026.
The headline writes itself. The timing is the part worth reading twice.
The collapse happened in 2025, before AI Overviews got big
Everyone assumes snippets died because AI Overviews ate them, and that the two curves crossed sometime this year. The monthly figures say otherwise.
| Month | AI Overviews | Featured snippets |
|---|---|---|
| Sep 2024 | 5.93% | 8.96% |
| Dec 2024 | 9.09% | 7.92% |
| Mar 2025 | 14.67% | 6.39% |
| Jun 2025 | 30.27% | 4.03% |
| Jul 2025 | 31.33% | 2.18% |
| Sep 2025 | 36.19% | 1.52% |
| Dec 2025 | 42.57% | 1.70% |
| Mar 2026 | 43.38% | 1.32% |
| Jun 2026 | 53.08% | 1.06% |
| Aug 2026 | 61.43% | 0.44% |
Snippets lost three quarters of their presence between September 2024 and July 2025, while AI Overviews were still only on about a third of results. The entire following year, during which AI Overviews doubled again to 61.43%, moved snippets from 1.52% to 0.44%.
So the feature was gutted in the first wave. What we have watched since is the tail.
That matters if you are still deciding when to change approach, because it means the decision point was eighteen months ago and the cost of waiting has already been paid.
The full 24 months, since nobody else has published it this way
| Month | AI Overviews | Local packs | Videos | Discussions and forums | Featured snippets |
|---|---|---|---|---|---|
| 2024-09 | 5.93% | 19.98% | 9.67% | 21.13% | 8.96% |
| 2024-10 | 6.81% | 20.11% | 10.78% | 23.07% | 9.52% |
| 2024-11 | 8.34% | 20.31% | 10.94% | 24.10% | 8.81% |
| 2024-12 | 9.09% | 19.89% | 10.02% | 24.93% | 7.92% |
| 2025-01 | 8.17% | 21.02% | 8.27% | 21.97% | 7.85% |
| 2025-02 | 7.84% | 21.56% | 8.06% | 14.67% | 7.45% |
| 2025-03 | 14.67% | 20.41% | 8.40% | 19.64% | 6.39% |
| 2025-04 | 17.63% | 17.51% | 6.69% | 15.61% | 5.36% |
| 2025-05 | 22.61% | 18.61% | 6.18% | 18.02% | 5.03% |
| 2025-06 | 30.27% | 19.40% | 6.01% | 17.56% | 4.03% |
| 2025-07 | 31.33% | 18.09% | 4.84% | 18.62% | 2.18% |
| 2025-08 | 30.70% | 20.50% | 5.85% | 16.11% | 1.92% |
| 2025-09 | 36.19% | 19.32% | 18.38% | 15.96% | 1.52% |
| 2025-10 | 43.76% | 11.47% | 31.05% | 14.36% | 1.47% |
| 2025-11 | 41.81% | 12.32% | 28.67% | 14.29% | 1.78% |
| 2025-12 | 42.57% | 16.54% | 28.35% | 13.13% | 1.70% |
| 2026-01 | 42.09% | 17.24% | 29.21% | 13.19% | 1.77% |
| 2026-02 | 42.79% | 26.76% | 25.32% | 13.11% | 1.57% |
| 2026-03 | 43.38% | 26.94% | 24.65% | 13.58% | 1.32% |
| 2026-04 | 45.00% | 25.98% | 24.93% | 17.02% | 1.04% |
| 2026-05 | 49.57% | 26.01% | 26.58% | 18.01% | 1.12% |
| 2026-06 | 53.08% | 26.34% | 26.27% | 11.95% | 1.06% |
| 2026-07 | 60.02% | 24.46% | 23.51% | 10.55% | 0.96% |
| 2026-08 | 61.43% | 25.99% | 19.98% | 7.42% | 0.44% |
Source: DataForSEO historical Google SERPs dataset, monthly panel, US English, 20 million keywords.
Three things in that table that the summary misses
Video did not double, it stepped. The summary figure is 9.67% to 19.98%. What actually happened is that video presence fell to 4.84% in July 2025, then jumped to 18.38% in September and 31.05% in October, then drifted back down to 19.98%. Two months did all the work, and video is currently falling from its peak rather than rising.
Local packs did something similar in reverse. They dropped to 11.47% in October 2025, which is the same month video spiked, then recovered to 26.76% by February 2026. If you run a local business and your map visibility fell off a cliff in late 2025 and came back, that was not you.
Discussions and forums have more than halved, from 21.13% to 7.42%. This is the one that cuts against the prevailing advice. The industry spent 2025 telling everyone that Reddit and forums were taking over the SERP, and the forum block specifically has been shrinking for two years. Reddit URLs are still turning up in nearly a third of results, which is a separate finding, but the dedicated discussions feature is going the way of the snippet.
Keep the method, change what you report
I am not going to tell you to stop writing answer-first content, because the method that won featured snippets is the method that gets a passage pulled into an AI answer.
Short answer in the first 40 to 60 words. A heading that names the question in the words a person would use. Comparisons in tables. Steps in ordered lists. Facts stated once, plainly, without three sentences of warm-up. Several vendors now report that pages which previously held a featured snippet are cited roughly twice as often in AI Overviews, which tracks with what we see, though those are vendor figures rather than peer-reviewed ones.
What changes is the scoreboard. There is no snippet box to screenshot in a monthly report any more. The equivalent evidence is whether your passage turns up inside the generated answer, and Search Console will not tell you, because it does not separate AI Overview data out.
We cover the mechanics of writing for retrieval in content chunking for AI retrieval. The short version is that a page which was easy to snippet is a page that is easy to cite, and the work transfers almost intact.
The caveat on all of this
United States, English language. Only the backlink section of DataForSEO’s report is global.
Snippets appear to persist longer in non-English markets and on query types where Google does not run an AI Overview, so a UK business should check its own Search Console before concluding that its own snippets have gone. We publish UK-specific figures separately for the same reason.
Where to start
Pull your top 50 queries out of Search Console and look at how many still return a featured snippet when you search them. For most businesses the answer will now be none, and that is a two-minute check rather than an argument.
Then decide what the snippet work becomes. If the pages are already answer-first, the job is measurement rather than rewriting. If they are not, the method has outlived the feature it was invented for, and it is worth doing now for a different reason.
If you want a view of where you are actually being cited rather than where you used to rank, the AI Visibility Score is free and takes a few minutes.