GEO Beyond SEO: How Mid-Market Companies Win Visibility in ChatGPT, Perplexity and Google AI Mode in 2026

Anyone searching on Google today increasingly gets a ready-made answer instead of a list of results. That changes how often anyone still clicks through to a website. An analysis by Pew Research based on the browsing data of around 900 adults in the US shows: when an AI summary appeared, users clicked on a traditional result in only 8 percent of visits, compared with 15 percent without a summary. They clicked on the sources linked in the summary itself in 1 percent of visits (Pew Research Center, July 2025).

For Germany, SISTRIX has quantified the consequences: AI Overviews appear for around 20 percent of search terms, costing German websites around 265 million organic clicks a month, roughly 6.6 percent of all organic clicks (SISTRIX, February 2026).

This article explains what Generative Engine Optimization (GEO) is, which levers are backed by studies, which widespread advice does not stand up to scrutiny and which steps make sense for mid-sized companies.

What is GEO - and where does classic SEO end?

Generative Engine Optimization (GEO) describes optimizing content so that it is mentioned and cited as a source in the answers of generative AI systems such as ChatGPT, Perplexity, Google AI Mode and AI Overviews, Microsoft Copilot, Gemini or Claude. The term comes from a study published in 2023 and presented at the KDD computer science conference in 2024 (Aggarwal et al., arXiv 2311.09735).

Classic SEO aims for a click from the results list. GEO aims for a mention in the answer. The two are not mutually exclusive. In its documentation on the AI features of Search, Google writes that the fundamentals of search engine optimization still apply there and that there are no additional requirements to appear in AI Overviews or AI Mode (Google Search Central). Other systems, however, select their sources according to their own patterns.

In the study that coined the term, methods such as adding quotations, statistics and source citations increased a piece of content's visibility in generative answers by up to 40 percent. The finding comes from tests with a purpose-built set of search queries. It shows a direction but guarantees nothing for an individual page.

What the data shows

Many figures circulate about AI search, and a good number of them cannot be found in the source given when checked. The following values can be traced in the linked sources:

1. Searches without a click: 59.7 percent of Google searches in the EU ended without a single click in 2024, 58.5 percent in the US (SparkToro with data from Datos, 2024).
2. Field experiment: in an experiment with 1,065 desktop users in the US, AI Overviews appeared for 42 percent of queries. For these queries organic clicks fell by 38 percent, and the share of searches without a click rose from 54 to 72 percent. The figures come from a working paper by Saharsh Agarwal and Ananya Sen from April 2026 (Search Engine Journal).
3. Germany: for search terms with an AI Overview, the click-through rate for the first position drops from 27 to 11 percent. While an average search leads to an organic click in 57 percent of cases, with an AI Overview it is only 33 percent. Depending on the industry, the click loss ranges from 1 to 24 percent (SISTRIX, 2026).
4. Cited or not cited: brands cited in an AI Overview achieved a 35 percent higher organic click-through rate on the same queries than uncited competitors. The basis is 3,119 informational search terms from 42 organizations between June 2024 and September 2025 (Seer Interactive, November 2025).
5. Quality of visitors: in HubSpot's State of Marketing Report 2026, 58 percent of marketers say that visitors arriving via AI tools convert better than classic organic traffic (HubSpot, 2026). That is self-reported, not measured.

The direction is clear: search terms with an AI answer bring fewer clicks, and those who appear in the answer lose less than those who are missing.

How the systems select their sources

The major answer systems draw on sources differently. For ChatGPT, Wikipedia is the strongest single source, for Perplexity it is Reddit, while Reddit hardly matters for Gemini. The frequently quoted percentages, however, usually refer to the ten most frequent sources rather than to all citations, and a mid-sized company finds it hard to get into either platform. What follows from this is described in the article Sources Behind AI Answers.

Google has one particularity: only a page that is in the search index and eligible to appear in Search with a snippet can serve as a source in AI Overviews or AI Mode. A page that is hard to find in normal search therefore rarely appears in the AI features either. Classic SEO remains the foundation.

The seven levers for mid-sized companies

1. The answer first

Start every main section with a concise, self-contained answer to the question the heading asks, and only then with the detailed explanation. AI systems prefer to adopt passages that make sense without the rest of the page.

Example: Instead of "In this article we will explain to you how AI agents work ..." → "An AI agent is software that completes tasks independently by connecting a language model with tools such as interfaces, database access and memory."

2. Clear structure

Each section should cover exactly one topic and be understandable without context from other sections. Headings that pick up the reader's question help people and machines alike to find the right passage.

3. Structured data: useful, but no citation lever

Schema.org markup makes information unambiguous: who the company is (Organization), who wrote an article (Person), when it was published and last changed (Article or BlogPosting) and which questions a page answers. There is, however, no proven citation advantage. Google writes that no special structured data is needed for AI Overviews and AI Mode, and an analysis by Search Atlas described by Search Engine Land found no correlation between schema coverage and citation rate (Search Engine Land).

4. Mentions outside your own website

AI systems mostly answer from third-party sources, not from the website of the company they are talking about. What works, therefore, is having the same information about services, region and target group in several independent places: in industry directories with editorial review, in trade media, on association and chamber websites and in review profiles. Contributions in forums and communities are only worthwhile where your own target group is actually active, and only as genuine contributions, not as advertising.

5. Authorship and experience

In December 2022 Google extended its quality criteria to include experience and has since spoken of E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness (Google Search Central Blog). In practice this means:

• Every article names a responsible person with a verifiable background.
• Experience becomes visible through concrete, traceable cases, not through claims.
• Figures and facts are backed by sources the reader can open.
• Reviews and mentions come from third parties, not from your own hand.

6. llms.txt: not a must

llms.txt is a proposal for a file intended to give language models a curated overview of a website, and it is often sold as an insider tip. Google, by contrast, states explicitly that no new machine-readable files or AI text files are needed for the AI features of Search. A lean, maintained file does no harm, but it does not replace any of the other levers.

7. Freshness and dating

Systems that search the web live can see when a page was published and last changed. Google recommends showing the date prominently on the page, labelling it as a publication or update date and also specifying it in structured data (Google Search Central). The update date should only change when the content has changed too.

What German SMEs need to add

AI in their own operations is now widespread among German companies: according to the ifo Institute, 54.4 percent of companies used AI in May 2026, up from 40.9 percent a year earlier (ifo Institute, June 2026). Whether AI systems describe your own company correctly to the outside world is a separate question that can only be answered by running your own queries.

Three points matter in addition for German SMEs:

• Transparency obligations: anyone who uses AI systems in customer contact must disclose this under Art. 50 of the AI Act; the obligation has applied since 2 August 2026 (article on the EU AI Act for AI agents). Clear notices also build trust.
• False information: if an AI system describes the company incorrectly, the cause usually lies in the sources. How to proceed is described in the article False Claims About Your Company in ChatGPT.
• Consistent master data: identical details on name, address and services on every site where the company appears prevent AI systems from mixing up locations or offerings.

Checklist: is your website ready for AI search?

• Every main section begins with a concise, self-contained answer.
• The headings are cleanly structured, and each section covers exactly one topic.
• Articles name a responsible person and show publication and update dates.
• Structured data for organization, author and article is present and error-free.
• The most frequent questions of the target group are answered visibly on the page.
• The robots.txt admits the crawlers of the AI providers you want to reach, or blocks them deliberately.
• Figures and facts are backed by sources that can be opened.
• Company name, address and service description match on all profiles and in all directories.
• The company is mentioned in trade media, by associations or in reviewed directories.
• A fixed set of questions is run regularly and repeatedly in the most important AI systems and logged.

Five common misconceptions

1. "GEO is just a new word for SEO." For Google's AI features, the same fundamentals as for Search do indeed apply. Other systems, however, select their sources according to their own patterns, and there mentions on third-party sites often count for more than your own website.

2. "An llms.txt is enough." According to its own documentation, Google does not require the file. At best it is a supplement.

3. "It won't work without a Wikipedia entry." Most mid-sized companies do not meet the inclusion criteria of the German Wikipedia. Consistent information in sources a company can actually reach is more effective.

4. "One query in ChatGPT shows where we stand." AI answers vary from query to query. A measurement only becomes reliable with fixed questions, several runs and logged results, as described in the article Measuring AI Visibility.

5. "Structured data is the most important lever." It makes information unambiguous, but a citation advantage has not been proven. Clear statements, evidence and mentions on third-party sites matter more.

Frequently Asked Questions (FAQ)

What is the difference between GEO, AEO and SEO? SEO (Search Engine Optimization) aims for good positions in the classic results list. AEO (Answer Engine Optimization) optimizes for direct answers such as featured snippets or AI Overviews. GEO (Generative Engine Optimization) is the umbrella term for optimizing to be mentioned and cited by generative AI systems. In practice AEO and GEO merge; SEO remains the technical foundation.

Do I need an llms.txt file? No. It does no harm, but Google states explicitly in its documentation that no new machine-readable files or AI text files are needed for AI Overviews and AI Mode. A clear page structure, unambiguous information and mentions on credible third-party sites matter more.

How do I measure whether I am cited in AI answers? With a fixed set of typical questions from your target group that you run several times at regular intervals in ChatGPT, Perplexity and Google AI Mode, logging the results with dates. Because AI answers vary, a single query says little; only repetition makes the result meaningful.

Do I lose clicks when AI Overviews appear for my search terms? As a rule, yes. SISTRIX measures for Germany that the click-through rate for the first position drops from 27 to 11 percent for search terms with an AI Overview. Being cited in the Overview itself limits the loss: Seer Interactive found a 35 percent higher organic click-through rate for cited brands than for uncited competitors on the same queries.

Do I need a Wikipedia entry or a presence on Reddit? Usually not. The German Wikipedia has inclusion criteria that most mid-sized companies do not meet, and Reddit is an important source only for some AI systems. For most companies, consistent information in industry directories, trade media and on association websites is more effective.

Does GEO also work for local SMEs? Yes, provided the information is unambiguous: a complete business profile on Google, consistent address and service details on every site where the company appears, and mentions in regional media or by associations. Contradictory information tends to produce vague or wrong descriptions.

How quickly do GEO measures take effect? There are no reliable timeframes. Systems that search the web live can pick up changes as soon as they fetch the page again; answers from a model's training data only change with its next version. Plan to observe for at least a quarter before evaluating a measure.

May I have texts written with AI? Yes. Google says it evaluates the quality of content, not how it was produced. Mass-produced pages without added value, on the other hand, fall under the spam policy on scaled content, whether people or machines wrote them. What counts are checked facts, sources and recognisable experience.

Conclusion: SEO remains the foundation, GEO comes on top

GEO does not replace SEO; it extends the discipline with a second layer: being mentioned in generative answers. The data shows that search terms with an AI answer bring fewer clicks and that cited sources lose less than missing ones. Mid-sized companies that give clear answers on their pages, keep their information consistent in independent places and measure their visibility repeatedly lay the foundation for being found in both worlds.

---

Sources

• Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., Deshpande, A. (2023, presented at KDD 2024): GEO: Generative Engine Optimization - arXiv 2311.09735
• Pew Research Center (22 July 2025): Google users are less likely to click on links when an AI summary appears in the results
• SISTRIX (25 February 2026): AI Overviews in Deutschland: So stark sinken die Klickraten wirklich
• SparkToro with data from Datos (2024): 2024 Zero-Click Search Study
• Search Engine Journal (2026): Report on the field study by Saharsh Agarwal and Ananya Sen on AI Overviews
• Seer Interactive (4 November 2025): AIO Impact on Google CTR: September 2025 Update
• HubSpot (2026): Answer Engine Optimization Case Studies, with figures from the State of Marketing Report 2026
• Google Search Central: AI features and your website
• Google Search Central: Influence your byline dates in Google Search
• Google Search Central Blog (December 2022): Update of the quality rater guidelines to include Experience
• Google Search Central Blog (February 2023): Google Search's guidance about AI-generated content
• Google Search Central: Spam policies for Google web search, section on scaled content
• Search Engine Land (2026): How Schema Markup Fits into AI Search - Without the Hype
• ifo Institute (5 June 2026): More than half of companies in Germany use artificial intelligence