Skip to content
← All posts
Artificial Intelligence· Updated

By VONA

GEO: How to Become Visible in AI Answers

Generative engine optimization in practice: how ChatGPT, Perplexity and Google AI overviews find sources and what you can do to get named.

More and more people no longer ask their questions on Google but on ChatGPT, Perplexity or the AI overviews of search engines. These systems answer directly and name the sources they rely on. For businesses that means: being found in a list of links is no longer enough — you want to appear in the answer itself. This article shows what is behind it, which measures have proven useful and how to check progress without fooling yourself.

What GEO is — and what it is not

Generative engine optimization, GEO for short, means preparing content so that AI-powered search systems find it, understand it and cite it as a source. It is not a replacement for SEO but builds on it: if you are technically clean, well linked and credible on your topic, you have better chances of being used by AI systems too. GEO adds to the question “How do I rank?” the question “How is my content understood, summarized and cited?”.

Important upfront: there are no official ranking factors for AI answers and no guarantee of being named. What this article contains are observations from tests and practice (as of October 2026). The systems change quickly.

How AI answers come about (simplified)

There are two ways a brand ends up in an AI answer:

  • Live research: the system searches the web, retrieves matching pages and has a language model write an answer from them. The principle resembles RAG. What counts here is whether your page is reachable, understandable and trustworthy.
  • Training knowledge: without research, the model answers from what it learned during training. That knowledge is limited in time. Anyone who is mentioned often and in the right context on the web has better prospects here in the long run.

Both point to the same foundation: good, easily findable content and a clear, consistent presentation of your brand everywhere.

What AI systems prefer in practice

  • Answer first: pages that pose a question clearly and answer it directly are easy to cite.
  • Structure: headings, short paragraphs, lists and unambiguous terms help machines organize content.
  • Specific instead of general: precise details, examples and sources seem more credible than platitudes.
  • Clear authorship: who is writing, who is behind it, what experience is there? Author information and an understandable “About us” page count.
  • Currency: a visible date and maintained content signal that information is not outdated.
  • Mentions across the web: references from other sites, industry directories and profiles strengthen credibility (see E-E-A-T: experience, expertise, authoritativeness, trustworthiness).

Step by step: GEO measures

1. Check access for crawlers. If AI providers may not retrieve your site, you cannot be cited there. Look into your robots.txt and the bot protection of your server or CDN. Well-known crawlers include GPTBot and OAI-SearchBot (OpenAI), PerplexityBot and ClaudeBot. Whether you allow all of them is a deliberate decision: blocking them protects content from training but may cost you visibility in answer search. Important: do not accidentally lock out the normal Google search (Googlebot).

2. Questions as headings, the answer right below. Phrase sections the way people ask (“What does … cost?”, “How does … work?”) and answer the question in the first sentence or two. Details, examples and exceptions follow.

3. Set structured data. With Schema.org markup (JSON-LD) you describe what a page is about: Organization and Person for providers and authors, Article or BlogPosting for posts, FAQPage for questions and answers, BreadcrumbList for the page hierarchy. Google rarely shows FAQ results as rich results today, but the markup still helps machines understand.

4. Make entities unambiguous. Use your company name, service names and contact details identically everywhere: on the website, in the imprint, in directories and profiles. The more clearly AI systems recognize who you are and what you stand for, the better they classify you.

5. Show evidence and currency. Link sources, state figures only with their origin and keep a visible publication and modification date. Revise important pages regularly.

6. Optionally add llms.txt. The file llms.txt in the root directory gives AI systems an overview of a website’s most important content. It is a proposal, not a recognized standard, and an effect on mentions is not proven. It is quick to create and costs nothing, though — we maintain one ourselves.

7. Build external presence. Maintain business profiles (for example Google Business Profile, industry directories), make sure you are mentioned in expert articles and directories and offer content that others like to link to.

Example: the same content, two structures (fictional)

Before: “Websites need good performance these days, and many factors influence it. Our experience shows that there are different approaches …”

After: heading “What improves a website’s load time the most?” — answer in two sentences: “Usually the biggest gains come from smaller images in modern formats, fewer and smaller scripts and a server that responds quickly. First, you measure where the time is lost.” — followed by a short list of the three most important levers with one example each.

The second version is more useful for people and for machines: the answer comes first, the content is structured, and each section can be cited on its own.

How to check progress

There is no standard metric for GEO. A simple approach has proven practical:

  1. Define 10 to 20 questions your target group looks for answers to.
  2. Ask them regularly (for example once a month) in ChatGPT, Perplexity and Google search.
  3. Note whether your brand is named, whether your page is linked as a source and which competitors appear.
  4. Check in your web analytics whether visits from AI services arrive. Google Search Console counts clicks and impressions from the AI features of Google search but does not report them separately.

Judge trends over several months, not single answers: AI answers vary from day to day.

Common mistakes

  • Writing texts only for machines: keyword lists and padded texts help neither readers nor AI systems.
  • Publishing AI texts unchecked: false statements damage credibility and may be spread further (see hallucination).
  • Locking out crawlers unnoticed: bot protection that blocks all unknown visitors also blocks AI crawlers.
  • Treating GEO as a one-off project: content, profiles and mentions need ongoing care.
  • Believing guarantees: anyone who promises a sure mention in ChatGPT promises something nobody can deliver.

Briefly answered

Does GEO replace search engine optimization? No. GEO builds on SEO and complements it.

Can I guarantee a mention in AI answers? No. You can improve the chances but not force them.

How long until something shows? That cannot be predicted reliably. Think in weeks to months, and check trends rather than single results.

Do I absolutely need an llms.txt? No, it is optional and a proposal, not a standard.

Conclusion and next steps

GEO is above all good editorial work with clean technology: clear answers, traceable sources, unambiguous details about you and a site that is easy to read for people and machines. Start with the first three steps — check access, answer questions, set structure and data — and then measure regularly. If you would like support, you will find our offer under SEO & GEO; you can look up terms in the AI glossary.

← Back to overview