In short
Almost everything published about how often Google shows AI Overviews is measured in English. We wanted to know whether the same holds in a market that doesn't search in English, so we ran 99 searches in Hebrew and recorded what Google returned. AI Overviews appeared in 36 of the 40 most popular queries we tested. The pattern held across commercial and informational intent, which matters for any business operating outside the English-speaking web.
Why measure a non-English market
If you only read English-language SEO research, it is easy to assume AI Overviews are a mostly-English phenomenon that will reach other languages eventually. For anyone running a business in a smaller language market, that assumption decides whether this is a problem for this year or a problem for later.
Israel is a useful test case: a small market, a right-to-left language with relatively little indexed content compared with English, and a population that searches heavily on mobile. If AI answers were going to lag anywhere, they would lag here.
What we did
We ran 99 queries in Hebrew across several categories - commercial searches with buying intent, informational "how do I" questions, local service searches, and product comparisons - and recorded whether Google returned an AI Overview, and whether the sources cited were local or international.
The 40 highest-volume queries in the set were treated separately, because those are the ones where the presence of an AI answer changes the economics of ranking.
What we found
- 36 of the 40 most popular queries returned an AI Overview. This is not an emerging trend in this market; it is the default state for popular searches.
- Informational queries triggered an overview more consistently than transactional ones. Searches phrased as questions almost always produced one.
- Local sources were cited regularly. Google was not falling back on English-language sources to answer Hebrew questions, which is what you might expect in a thin-content language.
- Where an overview appeared, the organic result below it was frequently the same domain. Being cited in the answer and ranking well were not in tension.
The limits of this study
We think the honest caveats matter more than the headline:
- 99 queries is a sample, not a census. A different query set would produce different percentages.
- Results were recorded at one point in time, from one location, on one device class. AI Overview presence varies by all three.
- Google personalises. Logged-out testing reduces this but does not eliminate it.
- We measured presence, not traffic impact. An overview appearing does not tell you what it did to clicks.
What it means if you operate outside English
The practical conclusion is that waiting is not a strategy. If AI answers are already the default on popular queries in a market this size, the same is almost certainly true of yours.
The pages that got cited in our set had something in common: they opened with a direct answer to the question, they pointed at an authoritative source where one existed, and the key facts were in the page text rather than in an image or a widget. None of that is exotic - it is the same work that makes a page good for readers.
Sources
Want to know what an AI agent sees on your site?
We check exactly that as part of an audit: crawler access, what is in the initial HTML, and the layer of information an agent can actually cite. Free SEO & AEO report → or talk to us →