AI Overviews appear on 0.9% of navigational queries and 57.9% of questions
AI Overview prevalence is 0.9% navigational, 2.1% transactional, 4.3% commercial, 13.1% branded against 24.9% non-branded. But the query forms content sites monetise — "X vs Y", "review", "best of" — measure 81-95%.
- How it was measured
- Ahrefs published a 146,122,391-SERP study with denominators, which is what makes it checkable. It independently replicates Pew's browsing-data study from a different sampling frame at 2,100x the sample: 1-word 9.5% vs Pew 8%, questions 57.9% vs Pew 60%.
- Evidence current as of
There is a real exemption from AI Overviews, and almost everyone describes it wrongly — including us, at first.
Prevalence by intent, from a 146-million-SERP study that publishes its denominators:
| Intent | AI Overview prevalence |
|---|---|
| Navigational | 0.9% |
| Transactional | 2.1% |
| Commercial | 4.3% |
| Branded | 13.1% |
| Non-branded | 24.9% |
| baseline | 20.5% |
And by query length: 1-word 9.5%, 7+ words 46.4%, questions 57.9%. That replicates Pew's independent browsing-data study (8%, 53%, 60%) from a completely different frame, six months later, at 2,100 times the sample. The length gradient is the best-established fact in this area.
The exemption protects the wrong thing
The protected classes are navigational, transactional and branded. Which sounds like good news until you notice what it means: a new publisher has no brand for anyone to navigate to. The exemption covers exactly the queries you cannot compete for.
Meanwhile the query forms a content site actually monetises measure, in a separate 49,358-query study:
| Query form | AI Overview saturation |
|---|---|
| "X vs Y" | 95.4% |
| "review" | 86.3% |
| "price / cost / buy" | 83.4% |
| "best of" | 81.3% |
And it is growing. Across 600,000+ keywords between November 2025 and April 2026, the share of commercial SERPs with an AI Overview grew 71% — monotonic, in all ten industries measured, with finance up 231%.
A methodological warning we earned the hard way
Our first read of this was wrong, and wrong in an instructive way. We quoted a vendor's own two-endpoint framing — commercial AI Overview share 8.15% to 18.57% — as evidence the exemption had closed.
The full thirteen-point monthly series shows something different: a single step discontinuity in April-May 2025, then seventeen months flat to declining, with commercial actually peaking in July 2025.
A two-endpoint comparison is structurally incapable of showing a discontinuity, and vendors publish endpoint framings because they are more dramatic than the series. The conclusion survived on better evidence. The reasoning did not.
Tracked in our ledger as EXEMPTION-IS-BRANDED, COMMERCIAL-IS-SATURATED, AI-OVERVIEW-CTR. If a number here is wrong, tell us and we will correct it in place and say what changed.