98% vs. 33%: Google's AI Overviews on brand searches in the US and in Germany

On 1 October 2026, the analytics company DemandSphere published figures on how often Google shows an AI Overview for brand-related searches: 26% on 1 September, a peak of 90% on 27 September and 82% on 29 September. Trade and business media picked this up, in Germany too. But the figures come from the US. Whether they also apply to the German market was open – there were no data on it.

On 6 October 2026, we searched Google for 40 internationally known brands – from Apple and IKEA to Lufthansa and Spotify, SAP and Siemens – in six countries: the US, the UK, Ireland, Germany, Austria and Switzerland. Each on desktop and on mobile, with six typical search patterns in the local language: brand name, “what is [brand]”, “[brand] reviews”, “[brand] alternative”, “[brand] prices” and, for subscription brands, “cancel [subscription]”. National brands and selected generic search terms without brand names ran alongside for comparison. In total, there were 7,962 non-personalised searches, queried within 16 minutes. For every AI Overview, we also recorded which sources it cites.

The countries were chosen so that language and EU membership can be told apart: English-speaking inside and outside the EU (Ireland, UK, US), German-speaking inside and outside the EU (Germany, Austria, Switzerland).

Across the search patterns tested, Google showed an AI Overview for 92% of brand searches in the US and 93% in the UK. In Ireland it was 56%, in Germany, Austria and Switzerland 51 to 52% each. So the UK behaves like the US, while the four other countries form a block of their own. But the average hides where the difference lies:

Questions that explain a brand are mostly answered with AI in the four comparison countries too: “what is [brand]” 84% of the time, “[brand] alternative” 72%, cancellation questions without exception. For searches shortly before a purchase, by contrast, the gap is large: for “[brand] reviews” and “[brand] prices”, the AI Overview appeared almost always in the US and UK, and in roughly one in three searches in Ireland, Germany, Austria and Switzerland. Some English price searches for listed brands are answered with the share price; without these brands, the gap remains large at 53 instead of 59 percentage points. Two repetitions on the same afternoon with ten of the brands show: the values fluctuate by a few percentage points between requests – for “reviews” in the US and UK between 90 and 98% – but the gap remains. A measurement by SE Ranking with 1,000 brand search terms (12 to 14 October) arrives at a similar gap with different search terms and tools: AI Overviews for 40% of brand searches in Germany and 83% in the US. For Germany alone, it looks like this:

Neither language nor EU membership explains the pattern alone: Switzerland is outside the EU and still at the level of Germany and Austria, English-speaking Ireland far below the UK. That fits a market-by-market rollout better. Google itself has not commented. Whether the DMA requirement that Google has had to implement in the EU since September 2026 also covers AI Overviews is still open (“Google warns of the worst search in 29 years”, in German).

Does Google treat brands in Germany more cautiously? To check, we queried the same search patterns with selected generic terms, such as “vacuum cleaner prices” or “running shoes reviews” – in the wording usual in each country. The gap also shows up in these comparable searches without brand names: for generic price searches, Google showed an AI Overview 96% of the time in the US and UK, and 34% in the four comparison countries. We find no measurable difference from brand searches.

“Many read the US figures as a forecast for Germany. For purchase searches, Germany is still far from that – not because Google spares brands here, but because the rollout has not got as far here. That points to a time lag rather than protection.”Dr. Michael Mohaupt, rethink.expert

Whether and when Google will expand AI Overviews here as in the US cannot be predicted from one measurement. If it happens, it would mainly affect the searches in which customers weigh their options. In addition, Google is testing showing AI Overviews expanded by default (“Who decides how much answer is on the results page?”, in German).

What the time lag should be used for is shown by a look at the sources of the AI Overviews. We analysed whether the first cited source belongs to the brand itself – that is, the source the answer relies on first:

Where the brand itself holds the authoritative answer – for cancellation, prices or terms, for example – it is often Google's first source. As soon as Google frames, rates or compares the brand with alternatives, control over the information increasingly shifts to external sources. For national brands, the gradient is the same, from 89% for cancellation questions to 9% for “alternative”. The middle stands out: for prices, the brand is the first source in only every second AI Overview; for “what is [brand]”, Wikipedia comes first more often (44%) than the brand itself (39%). For “what is SAP”, for example, Google's AI Overview in Germany cited Wikipedia, a YouTube video, a job portal and a training provider – but no SAP page.

What an AI Overview on a purchase search looks like is shown by “Lufthansa reviews”: Google answered with AI in all six countries on both devices – in eleven out of twelve cases without a single Lufthansa source. In Germany, the answer began: “Lufthansa receives a weak average of 1.3 out of 5 stars on Trustpilot, while the airline is rated more stably on Tripadvisor at 3.5 out of 5 stars.” “Frequent points of criticism” followed. Whether the picture is fair is not the question here. It shows in whose words a purchase search is answered.

“For authoritative information such as prices or cancellation, the brand is often Google's first source; for its reputation, almost never. Brands should therefore not ask whether AI talks about them, but on which questions they can be the authoritative source themselves – and where control over the information lies outside their own website.”Dr. Michael Mohaupt, rethink.expert

Which external places matter particularly also shows beyond brand search: in an analysis of 250,000 AI answers, almost one in three citations went to YouTube or Reddit (“Almost one in three AI citations”, in German). And anyone monitoring their own brand search should keep two things apart: whether the brand appears in the AI Overview – and whether visits and business result from it. How to measure this separately is described in “No right to the click” (in German).

The study runs for six weeks. The main report with the trend, full method, further findings and dataset is planned for mid-November.