How AI Overviews Work

AI Overviews in one sentence

AI Overviews are AI-generated summaries that Google displays directly above the classic search results. Instead of just delivering a list of links, Google summarizes the answer to a search query in its own words and names, below or beside it, the sources the answer is drawn from.

AI Overviews emerged from the Search Generative Experience (SGE), which Google launched as an experiment in 2023. In the US, AI Overviews were broadly rolled out in May 2024; in Germany, the rollout followed in spring 2025. They are therefore a fixed part of Google Search in the German-speaking region.

Which search queries trigger an AI Overview

Not every search shows an AI Overview. Google decides for each query whether an AI summary adds value. Roughly:

• Often triggered: informational questions, comparisons, how-tos, multi-part questions ("How does X differ from Y and which is better for Z?"). This is exactly where AI plays to its strength, because it bundles information from several sources.
• Less often triggered: purely navigational searches (someone is looking for a specific brand or URL) and simple transactional searches where the user wants to buy directly.
• Handled cautiously: sensitive topics in the areas of health, finance, or law (YMYL – "Your Money or Your Life"). Here Google is more reticent, because wrong AI answers can cause real harm.

Whether an AI Overview appears is also dynamic: Google tests continuously and adjusts the triggering with every update. The same search can show an AI Overview today and not tomorrow.

How an AI Overview is structured

A typical AI Overview consists of three parts:

1. The summary: one or more paragraphs of running text, often supplemented by bullet points or short subheadings. It answers the question directly.
2. The source references: linked cards or chips that show which websites Google drew the answer from. Users still reach your page via these links.
3. The expansion options: buttons or follow-up questions with which the user can dig deeper or switch into conversational mode.

Technically, behind the AI Overviews is Google’s in-house language model from the Gemini family, combined with the regular Google search index. The model does not invent the answer freely, but bases it on genuinely crawled web pages – a procedure that follows the principle of RAG (Retrieval Augmented Generation).

How Google selects the sources

For you, the central question is: by which signals does Google choose which pages it cites? Google does not publish the exact criteria, but reliable patterns can be derived from official statements and observable behavior:

• Classic ranking signals still count. AI Overviews draw on the same index as normal search. Pages that rank well for a query are more likely to be used as a source.
• A question is broken into sub-questions. Google uses a technique called "query fan-out": the search query is split into several sub-questions, and matching sources are sought for each. A page that answers one of these sub-questions particularly well can be cited – even if it is not in first place for the main keyword.
• Clear, extractable answers are preferred. Content that answers a question directly and precisely (answer first, then reasoning) is easier to adopt into a summary.
• Trust and authority matter. Trustworthy, topically established sources in the sense of E-E-A-T have better chances, especially on sensitive topics.

Example from the DACH region

If you search in Germany for "difference geo and seo", for example, Google often shows an AI Overview: a short paragraph that explains and contrasts both terms, plus three to four source cards from specialist sites. A user thus immediately gets the core message – and only clicks on a source if they want to go deeper. This is exactly the zero-click effect that the article The future of search examines more closely.

A second typical pattern is how-to questions like "how do I create a sitemap". Here the AI Overview often builds a numbered step-by-step list that merges individual steps from several sources. Anyone who wants to be one of these sources should provide the respective step in a self-contained and clearly worded form – because that is exactly how it gets adopted: as a single building block that can be lifted out of context.

What AI Overviews mean for your traffic

AI Overviews change not only how answers look, but also how much traffic reaches you. Two effects overlap:

• Fewer clicks on generic questions: When the AI Overview answers the question completely, the click is dropped for many users. Pages that only deliver generally known information are especially affected.
• More qualified clicks on complex topics: Anyone who still clicks after the AI summary has a deeper information need. This remaining traffic is more valuable, even as it shrinks.

The practical consequence: no longer measure your success by clicks alone, but also by whether your brand appears in the source cards. A mention in the AI Overview creates brand contact, even if no click follows.

How you get into the source list

Being cited in an AI Overview is the central goal of GEO. The full tactics are covered in the optimization chapters – here are the most important levers as a preview:

1. Write answer-optimized: answer the specific question clearly in the first paragraph before you elaborate.
2. Provide structure: use meaningful headings, lists, and tables. Machines extract structured content more easily.
3. Include facts and evidence: concrete numbers, data, and sources increase the likelihood of citation.
4. Be technically accessible: your page must be cleanly reachable and readable for Google’s crawler.
5. Build authority: establish yourself as a trustworthy source in your topic.

The concrete implementation for the AI Overview is explored in more depth in the article Optimizing for Google AI Overviews.

Conclusion

AI Overviews are Google’s answer to AI search: a summarized answer directly in the SERP, backed by the search index and a Gemini model. They appear above all on informational and multi-part questions, but are dynamic and are used more cautiously on sensitive topics. Anyone who wants to be cited as a source needs clear, well-structured, and trustworthy content – the same principles that also underpin classic SEO, only with an even stronger focus on direct, extractable answers.