The short version
- Question-form searches triggered an AI Overview 64.7% of the time against 9.5% for other searches, in a spring 2026 study of 55,393 trending US searches by Xu and colleagues.
- In our own test of 96 topics, the same topic showed an AI Overview 59.4% of the time as a short keyword and 93.8% as a natural question.
- Searches where Google showed a local map showed an AI Overview only 21.8% of the time in our study, against 81.1% without a map.
- Navigational searches, such as a brand or website name, returned an AI Overview only 12% of the time in a 2025 global test by Aral and colleagues.
- Overall rates range from 8.1% of fresh trending searches to 51.5% of representative real-user searches, so the mix of searches decides the headline number.
What is an AI Overview, and why does triggering matter?
An AI Overview is the AI summary Google places above its normal results for some searches, not all of them. If a search does not trigger one, there is nothing to be cited in, and the classic results page still decides who gets the click. When one does appear, it also affects whether people keep browsing afterward.
Google does not publish the rules. It says AI Overviews appear where AI adds benefit and are held back for “sensitive, explicit, rapidly evolving, or dangerous topics”, as Xu, Iqbal and Montgomery (opens in a new tab) summarize. Everything known about triggering therefore comes from researchers running large sets of searches and recording when a summary appeared.
Which kinds of searches trigger AI Overviews most often?
Questions and long, specific searches trigger them most; short keywords trigger them least.
Xu and colleagues ran 55,393 trending US searches over 40 days in March and April 2026. Searches that began with a question word showed an AI Overview 64.7% of the time, against 9.5% for the rest: a 6.8 times difference. Open-ended questions did best, with “how” at 84.3% and “why” at 73.4%. Simple lookups did less well, with “who” at 47.9% and “did” at 39.8%.
Length matters on its own too. Even among searches that were not questions, the rate climbed from 9.9% for single-word searches to 38.7% for searches of six or more words. A 2025 panel study by Pew researchers (opens in a new tab) found the same pattern in real searches by US adults: longer searches, searches starting with a question word and searches with both a noun and a verb tended to produce AI Overviews.
| Type of search | Showed an AI Overview | Study |
|---|---|---|
| Starts with “how” | 84.3% | Xu et al., trending US searches, 2026 |
| Any question word | 64.7% | Xu et al. |
| Starts with “did” | 39.8% | Xu et al. |
| No question word | 9.5% | Xu et al. |
| Complex “explain this” questions | 94.6% | Grossman et al., December 2025 |
| Amazon product keywords | 17.4% | Grossman et al. |
The last two rows come from Grossman and colleagues (opens in a new tab), who ran 11,500 benchmark searches in December 2025 from a simulated mobile phone in Newark, New Jersey.
Does the same topic get an AI Overview if it is worded differently?
Often yes: rewording a short keyword as a longer, specific question sharply raised the rate in two tests.
In our AI Overview frequency study, we searched 96 commercial topics three ways on 28 September 2026. As written, 59.4% showed an AI Overview; as a long non-question phrase, 87.5%; as a natural question, 93.8%. Most of the lift came from making the search longer and more specific, not from the question word itself. Every one of the 31 keywords containing a price or cost word showed an AI Overview.
Grossman and colleagues saw the same effect at larger scale. The same Natural Questions set triggered an AI Overview 86.2% of the time when phrased as questions, and 76.5% when cut down to keywords. Wording can also change which sources an AI answer cites.
Which searches rarely get an AI Overview?
Local map searches, brand and website searches, shopping keywords and breaking news rarely get one.
- Local searches. In our study, 21.8% of searches with a local map pack showed an AI Overview, against 81.1% without one. Searches containing “near me” showed one 17.3% of the time. Google seems to decide first whether a search needs a map or an explanation.
- Brand and website searches. In 2024, across seven countries, Li and Aral (opens in a new tab) found questions returned an AI answer 49% of the time, statements 16% and navigational searches 4%. In the 2025 rerun by Aral, Li and Zuo (opens in a new tab), the figures were 60%, 37% and 12%.
- Shopping searches. Only 5% of shopping searches returned an AI answer in 2024. By 2025 it was 13% worldwide.
- Breaking news. Grossman and colleagues found AI Overviews on only 8.1% of trending searches. The rate was 4.8% for topics trending for under five days and 12.7% after that.
Does the topic or industry matter?
Yes, topic and industry matter, but usually less than how the search is worded.
Huang and colleagues (opens in a new tab) ran 11,000 general-knowledge questions from one US location. Business questions produced AI Overviews 84% of the time, history 81% and education 74%. Sports (34%), technology (32%) and travel (32%) rarely did.
Xu and colleagues found a 13-fold spread across Google Trends categories, from 3.5% in beauty and fashion to 46.1% in hobbies and leisure. Politics (7.5%) and law and government (9.6%) sat below average, which fits Google’s stated caution on sensitive topics. Health (26.6%) did not, so that caution is not applied evenly.
Wording can override topic. Grossman and colleagues asked “What is the politician [NAME] known for?” for every member of the US Congress, and 93.8% produced an AI Overview. Politics is rarely summarized when the search is a trending name, but often when it is a direct question.
In our study of eight industries, B2B software keywords showed AI Overviews 96.0% of the time and financial services 88.0%. Legal, healthcare and franchise keywords were lower, largely because their results pages are dominated by local maps. Adjusting for that shrank the gap between industries from 60.0 to 36.7 points.
So how often do AI Overviews appear overall?
Between about one search in seven and two in three, depending on which searches are counted.
| Study | Searches counted | When | Showed an AI Overview |
|---|---|---|---|
| Xu et al. | 55,393 trending US searches | Spring 2026 | 13.7% |
| Chapekis et al. (Pew) | Real searches by 900 US adults | March 2025 | 18% |
| Wang et al. | Everyday searches of 1,100 US volunteers | March 2026 | 36% |
| Grossman et al. | Representative real-user searches | December 2025 | 51.5% |
| Underneath | 800 US commercial keywords | September 2026 | 60.8% |
| Aral et al. | Fixed set of benchmark searches, US | 2025 | 67% |
These figures do not contradict each other. Trending news searches rarely get a summary; the research-heavy keywords businesses track often do. Weighting by search volume also changes the answer: in our sample, the rate fell from 60.8% to 40.8%, because a few huge searches were mostly local or navigational.
The rate has also been rising. Running identical searches a year apart, Aral and colleagues found 67% of US searches answered by AI in 2025, against 42% in 2024. The figure of 36% comes from the field experiment by Wang and colleagues (opens in a new tab), a recent measure of what real people saw in everyday use.
What should you do about it?
Measure AI Overview exposure on your own keyword list, split by type of search, before planning around any average.
- Group your tracked keywords into questions and long phrases, short keywords, local searches, brand searches and shopping searches. Each group will behave differently.
- Treat “how” and “why” searches as the most exposed: in the Xu study, 84.3% of “how” searches showed an AI Overview.
- Track local searches separately. Where Google shows a map, your Google Business Profile still carries more weight than an AI summary.
- Recheck regularly. In our study, 83.3% of keywords had the same outcome two days apart, so roughly one in six changed.
- Do not read this as content advice. Question-shaped searches get more AI Overviews; no study here shows that question-shaped pages earn more citations. Whether you are cited once one appears is covered in Google rankings and AI Overview citations.
If you want help measuring this across your own searches, see our generative engine optimization service.
What does the research not tell us yet?
The research shows when AI Overviews appear, not why Google decides, and the rules keep changing.
- Google publishes no triggering rules, so every finding is observed from outside.
- Most studies use one US location and one point in time. Results by device, signed-in user or country are thinly measured.
- Query sets differ: trending searches, old question collections, commercial keywords. Only the Pew panel counts what real people typed, and its data is from March 2025.
- Several studies used AI models to label query types or topics, which adds some error.
- No study shows that changing a page changes whether it is cited once an AI Overview appears.
Frequently asked questions
What percentage of Google searches show an AI Overview?
There is no single figure. Studies range from 13.7% of trending US searches to 67% of a fixed US benchmark set, and Pew’s March 2025 panel of real searches found about 18%.
Do question searches trigger AI Overviews more often?
Yes, by a wide margin. Question-form trending searches triggered one 64.7% of the time against 9.5% for other searches, in a 2026 study by Xu and colleagues.
Do “near me” searches show AI Overviews?
Rarely. In our September 2026 study, searches containing “near me” showed an AI Overview 17.3% of the time, because Google usually shows a map instead.
Do brand searches show AI Overviews?
Rarely. Navigational searches, such as a company or website name, returned an AI answer 12% of the time in a 2025 global test, up from 4% in 2024.
Are AI Overviews becoming more common?
On identical searches, yes. Aral and colleagues found 67% of US benchmark searches answered by AI in 2025, against 42% in 2024.
Sources
- Xu, Iqbal and Montgomery (2026), Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact (opens in a new tab), arXiv:2605.14021.
- Grossman, Liu, Chen, Smith, Borcea and Chen (2026), How Generative AI Disrupts Search: An Empirical Study of Google Search, Gemini, and AI Overviews (opens in a new tab), arXiv:2604.27790.
- Chapekis, Lieb, Shah and Smith (2026), Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview (opens in a new tab), arXiv:2608.04831.
- Li and Aral (2025), Human Trust in AI Search: A Large-Scale Experiment (opens in a new tab), arXiv:2504.06435.
- Aral, Li and Zuo (2026), The Rise of AI Search: Implications for Information Markets and Human Judgement at Scale (opens in a new tab), arXiv:2602.13415.
- Huang and colleagues (2026), Answer Bubbles: Information Exposure in AI-Mediated Search (opens in a new tab), arXiv:2603.16138.
- Wang, Gleason, Bart, Wilson and Metaxa (2026), AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence (opens in a new tab), arXiv:2608.18352.
- Underneath (2026), When does Google show an AI Overview? 1,248 US searches