Google AI Mode: Independent from AI Overviews

Google AI Mode is not simply an extended version of AI Overviews. An Ahrefs analysis of 540,000 US query pairs from September 2025, in which AI Mode and AI Overviews cited the same URLs in only 13.7 percent of cases, shows that the two systems worked largely separately when selecting sources. Independent here means separate sources, not a separate product. For B2B companies in the DACH region, this has a strategic consequence: if you optimize only for AI Overviews, you have no guarantee of being cited in AI Mode as well. Since the launch in Austria, this is no longer a theoretical consideration. It is a concrete question of visibility.

What Google AI Mode is and why it is becoming relevant for B2B in the DACH region

Google AI Mode is a standalone AI search product within Google Search. Unlike AI Overviews, which appear automatically above the traditional search results, AI Mode is a dedicated mode: users switch into it and hold a conversational, multi-step dialogue with the search engine. Since January 2026, a follow-up question in an AI Overview leads directly into AI Mode in mobile search, and since I/O in May 2026, Google has merged both into a shared AI search interface. Technically, AI Mode works with a technique called query fan-out. The system breaks a single user query down into numerous sub-queries, answers them in parallel and combines the results into a detailed answer.

This depth shows in the data. According to the same Ahrefs analysis, AI Mode answers are around four times longer than AI Overviews and, with 3.3 entities per answer, name significantly more people and brands than AI Overviews with 1.3 entities. For your GEO strategy, this is a key insight: AI Mode offers more room for citations, more mentions of providers and therefore, in principle, more visibility opportunities per answer, but only for content that the system actually selects.

Since October 2025, AI Mode has been more than a fringe phenomenon in the DACH region. Google launched AI Mode on October 8, 2025, in 36 new languages and more than 40 new countries, including Germany, Austria and Switzerland. In Switzerland, it runs in German, French and Italian. According to Google, this makes it available in over 200 countries. Availability in German is decisive for Austrian companies, because it means that German-language industry queries are now routinely answered through AI Mode.

The market is ready. In Germany, according to the Bitkom study, 67 percent of the population use generative AI at least occasionally, compared with 40 percent the year before. Google Gemini, the model behind AI Mode, is used by 28 percent of respondents. At least in Germany, the user base for AI-powered search is therefore already broad.

The evidence of independence: why AI Mode chooses its own sources

The most important insight from the data available so far is that AI Mode's sources are independent of AI Overviews. The Ahrefs Brand Radar study, based on 540,000 US query pairs from September 2025, shows a citation overlap of only 13.7 percent. Even among the most prominent results on each side, the intersection remains small: if you look only at the top three citations of each, the overlap rises to just 16.3 percent.

In practice, this means that in roughly six out of seven cases, AI Mode draws on different sources than AI Overviews. A strong position in AI Overviews does not guarantee visibility in AI Mode, and vice versa. The cause lies in query fan-out. Because AI Mode breaks a query down into many sub-queries and looks for its own sources for each of them, it ends up with a broader and different set of sources than the more compact AI Overview logic.

This leads to a clear principle for your optimization strategy: AI Mode needs to be considered on its own. Optimizing a page for a single main query is not enough. Content has to be structured so that it covers the many possible sub-questions around a topic. That is exactly the lever that makes you eligible as a source in a fan-out system in the first place.

Gemini 3 Flash and 3.5 Flash: what the model switches change

AI Mode is updated continuously, and the underlying model is the most important factor for answer quality and citation behavior. Google made Gemini 3 Flash the default model in the Gemini app and in AI Mode in Search on December 17, 2025. The jump in reasoning capability is substantial. On the Humanity's Last Exam benchmark, Gemini 3 Flash scored 33.7 percent without tool use, while Gemini 2.5 Flash was at 11 percent. The next switch followed five months later: at its I/O developer conference in May 2026, Google made Gemini 3.5 Flash the new default model in AI Mode worldwide.

What does a stronger reasoning model mean for GEO in concrete terms? A model that breaks down complex tasks better and evaluates sources more precisely places higher demands on the depth and accuracy of the pages it cites. Superficial content tends to be selected less often as a reliable source, because the model is more likely to detect contradictions and thin evidence. Precise, well-structured and factually sound expert content gains a relative advantage.

One common misconception needs correcting here: the model switch alone does not make AI Mode independent of AI Overviews. The independence is demonstrated by the source divergence of 13.7 percent citation overlap, not by the model switch. Model switches change citation behavior and cause volatility, but they are no proof that the two products have been decoupled. The Ahrefs data also date from September 2025, before the switches to Gemini 3 Flash and Gemini 3.5 Flash and before the shared AI search interface of May 2026.

Concrete optimization tactics for AI Mode

GEO research provides an empirical basis for which content characteristics increase visibility in generative search systems. The widely cited GEO study by researchers from Princeton and other institutions shows that targeted optimization can increase visibility in generative answers by up to 40 percent. The study tested nine methods on a benchmark of 10,000 queries and identified several levers that transfer directly to AI Mode.

• Front-loading and answer islands: The core answer belongs at the beginning of a section. Since AI Mode looks specifically for extractable answers for each sub-query, you should create self-contained answer islands: a precise core sentence, followed by reasoning and evidence. This lets the system extract a single statement in isolation.
• Include statistics and quotes: For the strongest methods, the GEO study reports an improvement of 41 percent over the baseline on Position-Adjusted Word Count. Adding sources was particularly effective for lower-ranked pages: the Cite Sources method increased visibility for pages at SERP rank five by 115.1 percent. Evidence, figures and traceable sources are therefore a direct visibility factor, not an accessory.
• Entities and brand mentions: Because AI Mode answers name more entities than AI Overviews, a consistent entity strategy pays off. Your own brand should be clearly defined, linked to topics and mentioned across various trustworthy sources.

Structured data is not a separate lever for AI Mode: Google does not require any special markup and advises against fixating on it, but continues to recommend it as part of your general SEO strategy.

llms.txt is not a lever for AI Mode: according to Google's guide to generative AI features, Google Search ignores llms.txt files, so they have no effect on visibility in AI Overviews and AI Mode. If you maintain the file for other services, it does you no harm.

Using query fan-out as the core mechanism

If you want to understand AI Mode, you have to think in terms of query fan-out. The system does not answer one question but many small sub-questions at the same time. This calls for a different content structure than traditional SEO centered on a single keyword.

Instead of one page per main keyword, a topic cluster with semantic completeness is the better approach. A cluster covers a topic comprehensively and answers the obvious follow-up questions: what is it, how does it work, what does it cost, what alternatives are there, who is it suitable for. Each of these sub-questions should be designed as its own answer island with a clear heading and a compact, evidence-backed answer.

The 3.3 entities per AI Mode answer indicate how the system works: it relates several sources and players to each other. Content that defines terms cleanly, names players and makes relationships explicit fits this logic better than pure sales pages. For B2B in the DACH region, this means covering technical topics in their full breadth instead of only the aspects that are directly relevant to sales.

AI referral traffic and conversion: realistic expectations for B2B in the DACH region

For all the optimization, you need honest expectations about volume. AI search is still small in overall traffic. According to the international SE Ranking study of 101,574 websites, AI platforms accounted for only 0.32 percent of website traffic from January to April 2026, up from 0.24 percent in 2025 and 0.02 percent in 2024. For context: according to the same study, Google sends around 134 times more visitors than all AI platforms combined. Among the AI platforms, ChatGPT leads with 74.78 percent, followed by Gemini with 11.56 percent and Perplexity with 7.23 percent.

The decisive point for B2B is the quality of this traffic. AI visitors convert significantly better. According to the Semrush study, an AI search visitor is 4.4 times as valuable as a visitor from traditional organic search, measured by conversion rate. A second study confirms the direction: according to a study published by Search Engine Land, ChatGPT traffic converted at 1.81 percent, 31 percent higher than non-branded organic search at 1.39 percent.

The realistic reading for B2B in the DACH region is therefore: small volume, high quality. AI traffic today is no substitute for organic reach, but it is a valuable inflow of pre-qualified prospects. Its importance is growing: according to Semrush, AI search could deliver more visitors than traditional search for marketing and SEO topics by early 2028, and significantly earlier if AI Mode becomes the default.

Measurement and tools: tracking visibility in AI Mode

Measuring AI Mode visibility is demanding, because traditional tools reach their limits. Google Search Console reports clicks and impressions but does not cleanly break out AI Mode visibility as its own dimension. If you look only at GSC clicks, you systematically underestimate AI visibility, because a large part of the impact happens without a click.

A combination of citation and impression metrics therefore makes sense. Specialized tools such as Brand Radar track how often and with which sources a brand appears in AI answers. The analysis of 540,000 query pairs that demonstrates the source divergence was built on exactly this kind of data. The relevant metrics are:

• Citation share: How often is your own domain cited in AI Mode answers on relevant topics?
• Entity presence: Does your brand appear as one of the named entities? Given 3.3 entities per answer, this is an achievable target.
• Sub-query coverage: For how many of a topic's broken-down sub-questions does your own page appear as a source?

It is important to track these metrics alongside the traditional KPIs, not instead of them. AI visibility and organic traffic are separate balance sheets, and each needs its own targets.

Zero-click reality and KPI definition

The shift toward AI answers fundamentally changes the click logic of search. Generative answers resolve many questions directly on the results page, which lowers the click-through rate to linked content. The updated Ahrefs analysis based on Google Search Console data from December 2025 shows that the click-through rate for position 1 is around 58 percent lower when an AI Overview is present. That is nearly double the decline of 34.5 percent measured in April 2025.

Other data support this picture. According to the US analysis by Pew Research of 68,879 searches by 900 adults, users who saw an AI summary clicked a traditional search result in only 8 percent of cases, compared with 15 percent without an AI summary. Users clicked a link within the AI answer itself in only 1 percent of visits.

For your content strategy, this has a twofold consequence. First, your brand has to work even without a click: through a visible mention in the answer and a clear connection between brand and topic. Second, the KPI definition has to change. Pure click targets no longer match the new reality. Visibility and citation KPIs make more sense, combined with the underlying conversion quality, which the 4.4x factor documents.

Agentic commerce and the future: what lies ahead for B2B in the DACH region

AI Mode is evolving from a pure answer engine into an assistant that takes action. Agentic functions such as deeper research, bookings or completing multi-step tasks shift the focus from information to action. For retail, this creates agentic commerce scenarios in which the AI system compares products and initiates transactions.

For B2B lead generation in the DACH region, a clear pattern is emerging over the medium term. As AI systems increasingly do the preliminary research and build shortlists, presence in exactly these AI answers decides whether a provider makes the shortlist at all. The measured source divergence of 13.7 percent from September 2025 suggests that this presence in AI Mode has to be built separately. If you are present here early with cleanly structured, evidence-backed expert content, you secure a place in your customers' AI-supported selection processes.

Further reading

Based on the data available so far, Google AI Mode chooses its own sources, which makes it a separate optimization channel rather than an appendage of AI Overviews. The key levers are topic clusters built for query fan-out, front-loading with answer islands, evidence-backed statistics and a consistent entity strategy. If you want to build visibility in AI Mode, explore the related GEO topics in parallel: how query fan-out works, how to build answer islands, the Princeton lever of statistics and quotes, and the measurement of AI visibility with citation and impression metrics. What matters is to keep AI visibility and traditional SEO as two separate but complementary balance sheets.