How to check if your brand appears in Google AI Overviews

AI Overviews are the machine-generated summaries Google places above the normal results for some searches. Instead of ten blue links, the searcher gets a few paragraphs of synthesised answer, assembled by Google's Gemini models from pages across the web, with citation links tucked alongside or behind expandable chips. For many searches they are now the first thing people read, and sometimes the only thing.

Google does not publish the rules for when an Overview triggers. The observable pattern is that they appear most often on question-shaped and research-shaped queries: how-to searches, comparisons, 'best X for Y', 'is X worth it'. They appear far less often on navigational searches (someone typing your brand name to find your site) and on purely transactional ones, where Google tends to show shopping results instead. That distribution matters, because the queries most likely to trigger an Overview are precisely the ones where buyers are still deciding, the moment when being named or ignored changes outcomes.

Start by writing down ten to twenty queries a real buyer would use. Not your brand name but category queries: the problems you solve, the comparisons you show up in, the 'best' lists you want to be on. If you sell project management software to UK agencies, the list looks like 'best project management tool for agencies', '[your product] vs [competitor]' and 'how to choose project management software'.

Then check them properly. Open a private or incognito window, log out of Google, and run each query. For every search, note three things: whether an AI Overview appears at all; whether your brand is named in the text of the answer; and whether any of your pages sit among the cited sources. Click the link icons or expand the citation panel, because sources are often hidden behind it.

Location and repetition matter more than people expect. Overviews differ by country and sometimes by city, so if you sell beyond your own region, check from there too; a VPN, or the region option in Google's search settings, gives you a rough view. Then run the same queries again a few days later. Google is constantly experimenting: whether an Overview appears, what it says and whom it cites can all change between two searches that look identical.

There is no such thing as 'the' AI Overview for a query. What you see is one generation, for one user, in one place, on one day. Treat any single screenshot as a sample, not a verdict.

Because Google does not publish trigger criteria, the honest approach is to test shapes of query and see what happens in your category. In our experience, the shapes most likely to produce an Overview for commercial topics are 'best [category] for [audience]', '[option] vs [option]', 'is [product type] worth it', 'how to choose a [category]' and 'alternatives to [market leader]'. Price-led and purely transactional queries, such as 'buy running shoes size 9', trigger them far less often.

Build your checking list from those shapes, seeded with your own category terms. If none of your twenty queries produces an Overview at all, that is a finding too: this surface is not yet where your buyers are being influenced, and you can put the effort into the AI assistants they do use instead.

An AI Overview is built from cited pages, so getting in is about earning citations rather than simply ranking. The two overlap, because pages that rank well are more likely to be drawn on, but they are not the same thing. Overviews frequently cite a page that answers one sub-question crisply over a bigger page that ranks higher but answers vaguely.

If your brand is absent, three moves help. First, make sure a page on your site answers the target query directly, in plain language, in its first paragraph, because Overviews lift from pages that state the answer, not pages that build up to it. Second, cover the comparison and 'best for' angles honestly, including where you are not the right fit; synthesised answers reward specificity. Third, look at who is being cited for your queries. Often it is independent reviews, industry publications and listicles rather than any vendor's own site. For 'best X' queries especially, getting your brand into the third-party pages Google already draws on can matter more than anything you publish yourself.

None of this is instant. Citation changes tend to follow content changes with a lag, which is another reason a one-off manual check cannot tell you whether the work is paying off.

Manual checking is fine to start with, but you will outgrow it quickly. It does not scale past a couple of dozen queries, it cannot separate noise from real change, and every screenshot goes stale the moment you take it. The variability that makes Overviews hard to check once is exactly what makes tracking them a repeated-measurement problem: the same queries, run regularly, from consistent conditions, with mentions and citations logged over time. Our Google AI Overviews tracker page covers this in detail, including what can be measured reliably and what genuinely cannot.

It is also worth widening the lens. Google is one surface among several, and AI systems get brands wrong more often than most marketers assume: in our July 2026 test of 30 brands, 27 were described with at least one materially wrong claim when the AI answered without live web search (full study). You can see what an AI assistant currently says about yours in a couple of minutes with our free AI visibility checker. It tests Claude today, with other engines rolling out. Either way, do the manual checks above this week. They cost an hour, and they will tell you whether AI Overviews are already shaping how buyers first meet your brand.