Getting Recommended by AI: What GEO Is and How It Works
More and more people no longer ask Google but an AI, and get an answer instead of a list of links. If you're not in that answer, you don't exist for that customer.
Generative engine optimization (GEO) covers all the measures that prepare content so that AI systems such as ChatGPT, Google AI Overviews, Perplexity or Claude cite it in their answers. The key difference from classic SEO: it's no longer about ranking first in a list of results, but about being named as a source in a finished, written answer.
The good news first: GEO isn't a new discipline that makes everything before it worthless. The overlap with good SEO is around 80 percent. The remaining 20 percent, however, decides whether you're cited or merely read.
Why this is changing right now
Google answers more and more searches directly in its AI Overviews. Users get a summarized answer with sources and often don't click any further. For information-seeking searches, click-through rates have dropped noticeably as a result.
At the same time, people increasingly do their research directly in chat tools. Someone looking for an IT provider in the Frankfurt area may no longer ask Google, but type the question into ChatGPT. The answer names three companies. Whether yours is one of them decides who gets the first contact.
The most important difference: with Google, ranking mattered. With AI answers, being quotable matters. A page can rank eighth and still be cited if it answers the question more precisely than the pages above it. Studies show that a considerable share of the pages cited in AI Overviews don't come from the top five positions.
What AI systems cite and what they don’t
AI systems extract blocks of text that can stand on their own. A section gets cited when it meets three conditions:
- It answers the question completely, without the reader needing the rest of the page.
- It contains the statement and the reasoning close together.
- It's written unambiguously, with a clear subject and no references like "as described above".
What doesn't get cited is the classic structure of many corporate texts: a long-winded introduction, then three paragraphs of context, and the actual point somewhere in the last third. That was good enough for search engines, but it's useless for AI extraction.
Seven measures that work
1. Start with the answer
Every page and every major section starts with the direct answer to the question in its heading. The context comes after. This is the single most effective measure, and the one writers find hardest.
2. Structure with questions
Subheadings that match real search queries: "How much does managed IT cost per month?" instead of "Pricing models". The reason: AI systems break complex queries into sub-questions and look for a matching passage for each one.
3. Be specific
Numbers, timeframes, prices, thresholds. "Starter from a €49 base price plus €39 per device per month" is quotable; "attractive terms" isn't. If you give numbers, you become the source; if you give adjectives, you get passed over.
4. Bring in your own data and experience
AI systems can phrase general knowledge themselves; they don't need a source for that. What gets cited is what they can't produce on their own: your own figures, concrete cases, observations from your industry. A sentence like "In our projects, moving a company's IT typically takes two to three weeks" is valuable for exactly that reason.
5. Make authorship transparent
Google has now explicitly emphasized the importance of author information in its own documentation. Articles with named authors and a verifiable background, a genuine about page and complete legal information are trust signals that feed into the selection.
6. Use structured data
Markup based on schema.org helps systems classify content: organization, article, author, local presence. Keep your expectations right, though: structured data doesn't force a citation. It's a supporting signal; the visible text does the main work.
7. Get mentioned beyond your own website
AI systems cross-check sources against each other. If you appear in industry directories, expert articles, local press and review platforms, you count as an established player. A website on its own, with no mentions anywhere else, looks to these systems like an unsupported claim.
What explicitly doesn’t work
Treating FAQ markup as a miracle cure. Google has now completely discontinued the expandable FAQ results in search. The markup itself is still useful to aid understanding, but it doesn't trigger citations.
Adding a file like llms.txt and hoping. Google has explicitly stated that such files don't get special treatment.
Neglecting classic SEO. AI systems draw on the same index as regular search. If you abandon the fundamentals, you lose both at once.
Stuffing texts with keywords. That stopped working with Google long ago, and it works even less with language-processing systems.
How to measure whether it’s working
The classic metrics are no longer enough. Add these:
- Manual spot checks. Regularly ask your most important search questions in ChatGPT, Perplexity and Google yourself and see who gets named. It's the most direct method.
- Check your traffic sources. Visits from chat services and Perplexity show up as separate sources in web analytics. The numbers are still small, but they're growing.
- Watch Search Console. Falling click-through rates with stable impressions are a typical sign that the answer is already being given in search.
What this means for smaller businesses
Some good news to finish: for small and mid-sized businesses, the situation is more favorable than with classic SEO. For generic search terms, you're competing with corporate budgets. For concrete, specific questions, such as "How much does managed IT cost for a law firm with twelve workstations?", whoever answers precisely and honestly wins.
And a small business can often give that precise answer better than a corporation, because it actually knows the case.
Frequently asked questions
What is GEO (generative engine optimization)?
GEO covers the measures that prepare content so that AI systems such as ChatGPT, Google AI Overviews or Perplexity cite it in their answers. Unlike classic SEO, it's not about ranking in a list of results but about being named as a source in a written answer.
What's the difference between SEO and GEO?
SEO aims for a good ranking in the list of results; GEO aims to be cited in an AI-generated answer. The overlap is around 80 percent. The main difference is in how text is structured: GEO calls for direct, self-contained answers at the start of every section.
How do I get into Google's AI Overviews?
By writing sections that answer a question completely and on their own, using concrete numbers instead of adjectives, structuring headings as questions, making authorship transparent and getting mentioned beyond your own website.
Is FAQ schema still useful for search engine optimization?
Google has completely discontinued the expandable FAQ results in search. The markup is still useful to help search engines understand content, but it doesn't trigger citations. The visible structure of the text on the page does the real work.
Does GEO replace classic search engine optimization?
No. AI systems draw on the same search index as regular Google search. GEO complements classic SEO but doesn't replace it. If you neglect the technical and content fundamentals, you lose visibility in both.
