Business owners often tell me something like this: “AI Overviews are everywhere now, so our traffic is doomed.” But when you actually look at a keyword list, it often turns out that most of the money keywords don’t show an AI Overview at all. Other sites find the opposite: their whole blog sits on queries that almost always trigger one.
The truth is that AI Overviews are not spread evenly. Google shows them for some kinds of searches far more than others. If you know which kinds, you can map your exposure, plan content and set realistic traffic expectations instead of guessing.
Let me explain what the largest study so far found about AI Overview triggers, what Google itself says, and how I suggest you sort your own keywords into high, medium and low exposure. This guide is about when an AI Overview appears. If you want to know how to get cited once one does appear, see my guide on getting your content cited in Google AI Overviews.
How the 146-million-SERP trigger study worked
The main data here comes from Ahrefs. In What Triggers AI Overviews? (November 2025), Ryan Law and Xibeijia Guan analysed 146,122,391 desktop search results pages from September 2025, tagged with 86 keyword traits, such as whether the query was a question, its intent, whether it was branded and its topic category.
“AI Overviews appear on 20.5% of all SERPs (29,991,998 out of 146,122,391).”
— Ryan Law, Ahrefs, What Triggers AI Overviews?
That 20.5% is the baseline. Every number below is more useful when you compare it to that figure. A trait with a 40% trigger rate makes an AI Overview about twice as likely as average. A trait with a 4% rate makes it about five times less likely.
What this study can’t tell you
Before we go further, a few honest caveats. The data is desktop only and from one month. The keyword labels (intent, category, YMYL) come from Ahrefs’ own classifiers, which won’t be perfect. And Google changes how often it shows AI Overviews over time. So I treat these numbers as a strong map of tendencies, not as fixed rules for your specific keywords.
Google’s own description is short. Its page on AI features and your website says: “AI Overviews are only shown when our systems determine that it is additive to classic Search.” The same page notes that they often don’t trigger. The data below is the best picture we have of what “additive” means in practice.
Do question-style and longer searches trigger AI Overviews more?
Yes, by a wide margin. This is the clearest pattern in the whole study.
- Question queries (who, what, why, how, can, does…) triggered AI Overviews on 57.9% of results, compared with 15.5% for non-questions.
- Longer queries triggered them far more often. The rate climbed steadily with the number of words.
| Words in the query | AI Overview rate (Ahrefs, Sept 2025) | Compared with the 20.5% baseline |
|---|---|---|
| 1 word | 9.5% | Less than half as likely |
| 3 words | 14.0% | Below average |
| 5 words | 27.6% | Above average |
| 7+ words | 46.4% | More than twice as likely |
A separate data set points the same way. Pew Research Center tracked the real browsing of 900 US adults in March 2025. As reported by GIGAZINE’s summary of the Pew study, AI summaries appeared on about 8% of searches of one or two words, but on 53% of searches with 10 or more words.
Why would this be? A short query like “running shoes” is usually best served by a list of shops and products. A long one like “why do my knees hurt after running on a treadmill” is a real question that needs explaining. That is exactly where Google seems to decide a written summary is “additive.”
Informational, commercial and transactional searches: the biggest divide
Intent is the goal behind the search. If you need a refresher, my guide to search intent explains the four main types. For AI Overviews, intent creates the sharpest split of all.
| Intent | Example | AI Overview rate | Share of all AIO keywords |
|---|---|---|---|
| Informational | “how does a heat pump work” | 21.4% | 99.9% |
| Commercial | “best heat pump brands” | 4.3% | 5.5% |
| Transactional | “buy heat pump online” | 2.1% | 1.2% |
| Navigational | “daikin login” | 0.9% | 0.1% |
Source: Ahrefs, September 2025 data. Shares add to more than 100% because a keyword can carry more than one intent label.
“Google seems unlikely to trigger AIOs for other intent types.”
— Ryan Law, Ahrefs, What Triggers AI Overviews?
This is good news for many businesses. If most of your revenue comes from commercial and transactional searches (“plumber near me,” “buy standing desk,” “best CRM for small teams”), your money pages are much less exposed than the headlines suggest. Your blog and guide content is where the exposure sits.
The same study also looked at Google’s own “Know / Do / Website / Visit-in-person” framework. “Know” queries triggered at 21.4%, “Do” queries at 13%, visit-in-person queries at 7.1% and website queries at just 0.9%.
Branded, navigational and local searches
Searches that contain a brand name triggered AI Overviews less often: 13.1% for branded keywords versus 24.9% for non-branded. Non-branded keywords made up about three quarters (76.3%) of all AI Overview keywords.
Local searches were even lower. Queries with local intent triggered AI Overviews on 7.9% of results, compared with 22.8% for non-local ones. That makes sense, because local searches already have the map pack, which answers the question with real businesses.
One important nuance on brands. “Branded” covers a lot: “nike,” “nike returns policy” and “is nike ethical” are all branded but behave very differently. Questions about a brand can still trigger AI Overviews. I cover what to do when they appear for your brand in AI Overviews for brand searches.
YMYL topics: why health and money queries trigger more, not less
YMYL stands for “Your Money or Your Life.” It is Google’s term for topics where bad information could harm someone’s health, finances, safety or wellbeing. Many people expected Google to be cautious about AI answers here. The data says the opposite.
- YMYL queries triggered AI Overviews on 34.3% of results, versus 17.2% for non-YMYL.
- Medical queries were highest at 44.1%.
- Safety (31%), legal (23.6%) and financial (22.9%) were also above or around the baseline.
- By broad category, Science (43.6%) and Health (43.0%) led the whole study.
The Ahrefs authors called this surprising given the stakes, and I agree. In practice, it means health, legal and finance publishers carry some of the highest exposure. If you work in these areas, my guide on YMYL topics and algorithm updates explains why these niches also swing harder in core updates.
Query shapes that pull AI Overviews in, and ones that keep them out
Beyond intent, Ahrefs classified queries by their “shape.” Some shapes pulled AI Overviews in strongly:
- Reason queries (“why is my…”) had the highest rate in the study at 59.8%.
- Yes/no queries (“can dogs eat grapes”) were at 57.4%.
- Definitions (47.3%) and consequences (“what happens if…”, 46.7%) were also high.
- Comparisons were lower than I expected, at 26.2%, only a little above baseline.
Other kinds of query kept AI Overviews out:
- Breaking news: very newsy keywords triggered only 6.3% of the time. AI Overviews lean towards evergreen topics.
- Shopping (3.2%) and real estate (5.8%) were the lowest categories.
- Adult and gambling queries were around 1.5% or lower.
What about other SERP features?
The Ahrefs study did not test SERP features like shopping ads, map packs or video carousels as triggers, so I can’t give you numbers. But the low rates for shopping, local and news queries line up with where Google already has strong special features (product listings, the local pack, Top Stories). My reasoning, which I treat as a hypothesis rather than a finding, is that when Google already has a purpose-built feature that answers the search well, an AI summary is less “additive.” If you want more on how features compete for space, see what happened to featured snippets in the AI era.
Mapping your own keywords by AI Overview likelihood
Here is where this becomes useful. Instead of guessing, you can sort your keyword list into exposure tiers. Here is how I would do it.
- Export your keywords. Pull the queries from Search Console (last 3 months) or your rank tracker, with clicks and impressions.
- Check actual AI Overview presence where you can. Many rank trackers and SEO tools now flag whether a SERP shows an AI Overview. Use real data first; use the rules below only for keywords without it.
- Tag each keyword with simple traits: question (yes/no), word count, intent, branded (yes/no), local (yes/no), YMYL (yes/no).
- Score the likelihood. Use the matrix below to place each keyword in high, medium or low exposure.
- Weight by value. Multiply by clicks or conversions so you see where exposure overlaps with revenue, not just traffic.
- Decide an action per tier, as in the last column.
| Exposure tier | Typical traits | Expected AIO likelihood (from the data above) | What I would do |
|---|---|---|---|
| High | Question, 5+ words, informational, YMYL or science/health | Roughly 40–60% | Aim to be cited; write answer-first; expect lower click-through |
| Medium | Informational but short, non-question, or comparison | Roughly 15–30% | Monitor; strengthen unique angles; track click-through monthly |
| Low | Commercial, transactional, branded navigational, local, news, shopping | Under 10% | Classic SEO and conversion work; low AI Overview risk for now |
Example: a home services company’s 200 keywords
This is a hypothetical example with made-up numbers. A heating and cooling company exports 200 keywords bringing 6,000 organic clicks a month. After tagging:
- High exposure: 70 keywords (blog questions like “why is my heat pump blowing cold air”), 2,100 clicks, almost no direct leads.
- Medium exposure: 50 keywords (“heat pump vs furnace,” “heat pump sizes”), 1,500 clicks, some leads.
- Low exposure: 80 keywords (“heat pump installation [city],” “emergency AC repair near me”), 2,400 clicks, most of the leads.
Now say the company’s high-exposure pages lose about a third of their clicks when an AI Overview appears. Ahrefs’ click-through study estimated a roughly 34.5% drop in click-through for the top-ranking page when an AI Overview was present (April 2025). That would mean around 700 fewer blog clicks a month, which sounds scary. But because the money keywords sit in the low tier, the lead impact is small. The business should protect its service pages, and treat the blog as a citation and brand-building play. I show how to estimate this per keyword in how much traffic AI Overviews take.
What to do with pages in the high tier
High exposure doesn’t mean “give up on the page.” It means the page’s job changes. It should be the clearest, most trustworthy answer on the topic, so it is the one Google’s AI cites. Google’s guidance on creating helpful, reliable, people-first content is the right checklist here: show who wrote it, why they are qualified, and add something original, such as your own photos, prices, data or worked examples, that a summary can’t replace. Then add a clear next step on the page, such as a calculator, a booking link or a related service, for the readers who do click through.
Tools for this step: your keyword map is the natural home for an “AIO tier” column, and my guide on AI-proof keywords explains how to find queries that AI answers can’t fully satisfy.
How AI Overview triggers are shifting, and how to keep up
Everything above is a snapshot. Google has expanded and adjusted AI Overviews many times since they launched, and AI Mode now offers a separate, fuller AI experience. Google says AI Mode is especially useful for queries that need more exploration, reasoning or complex comparisons. I compare the two in AI Mode vs AI Overviews.
What I would expect to keep changing:
- Branded and commercial coverage. These were low in September 2025 data, but this is exactly where Google has the most room to expand.
- Mobile versus desktop. The Ahrefs data is desktop only. Check mobile results for your key terms separately.
- Measurement. Google says clicks from AI features are included in Search Console under the Web search type, and its generative AI guide points to a generative AI performance report for measuring how content performs in these features. My guide on what you can and can’t see about AI Overviews in Search Console explains the limits.
My suggestion: re-check AI Overview presence for your top 100 keywords every quarter, and move keywords between tiers when the SERP changes. It takes an hour and keeps your forecasts honest.
Questions about what triggers AI Overviews
What percentage of Google searches show an AI Overview?
In Ahrefs’ September 2025 data, 20.5% of 146 million desktop results pages showed one. The rate varies hugely by query type, from under 1% for navigational searches to nearly 60% for “why” questions.
Do commercial keywords trigger AI Overviews?
Rarely. Commercial keywords triggered them 4.3% of the time and transactional keywords 2.1%, compared with 21.4% for informational keywords.
Do AI Overviews show for local searches?
Much less often. Local queries triggered AI Overviews on 7.9% of results, versus 22.8% for non-local queries.
Why do health queries get so many AI Overviews?
Health queries are often long, informational questions, which is exactly the type that triggers AI Overviews most. Medical YMYL queries triggered them 44.1% of the time.
Can I stop an AI Overview from showing for my keyword?
No. Google decides when to show one. You can control how your own content is used in snippets with tags like nosnippet, but that affects your snippets in Search too.
Where to start mapping your AI Overview exposure this week
If you only have an hour, export your top 100 keywords from Search Console, tag each one as question or not, and as informational, commercial or transactional. That alone will show you roughly where your exposure is. Then add clicks and leads, and you will know whether AI Overviews are a traffic issue, a revenue issue, or neither.
For the pages that sit in the high tier, focus on being the source that gets cited. My guides on whether content length matters for AI citations and which citation sources you can influence are good next reads. If you would like us to map your keywords and build a plan around it, our SEO services and local SEO team can help. Get in touch for a free quote.
Further reading on what triggers AI Overviews
- Which AI Citation Sources Can You Actually Influence?
- Which Search Index Does ChatGPT Use? Google, Bing and Its Own Crawler
- AI SEO Statistics: Key Numbers on AI Search and Citations
Sources for this AI Overview triggers guide
- Ahrefs — What Triggers AI Overviews? 86 Factors and 146 Million SERPs Analyzed
- Google Search Central — AI features and your website
- Google Search Central — Optimizing for generative AI
- Ahrefs — AI Overviews reduce clicks
- GIGAZINE — Report on Pew Research Center’s AI summary study
- Google Search Central — Creating helpful, reliable, people-first content
