GEO Optimization: How to Get Named in AI Answers

At a glance

What

You optimize your website and your details on the web for answers from AI systems.

Origin

The term comes from a research paper from 2023, presented at the KDD 2024 conference.

Foundation

Clear facts, clean structure and identical details about your business everywhere.

Complements SEO

Google says the basics of search engine optimization are also decisive for AI features.

Technology

Each provider has its own crawlers, which you control separately in robots.txt.

No guarantee

The providers of the AI systems decide for themselves what they name.

This guide explains what GEO optimization means, how AI systems find their sources and what you do in practice. At its core is a checklist with 12 steps, plus a prompt test for measuring and the most common misconceptions.

GEO optimization covers everything that makes sure AI-powered search systems know your business, describe it correctly and name it for relevant questions. Instead of a list of ten links, these systems deliver a finished answer. Whether you appear in it depends on which sources they find and how clear those sources are.

Researchers around Pranjal Aggarwal coined the term Generative Engine Optimization in a paper that appeared in November 2023 and was accepted at the KDD conference in 2024. They built their own test dataset for it, called GEO-bench. In their experiments, the visibility of sources in AI answers rose by up to 40%. That is a lab result with an upper limit, not a promise for your website.

GEO, AEO and SEO: three names, one foundation

Besides GEO, you also read AEO, Answer Engine Optimization. It stands for the same goal: appearing in answers instead of only in result lists. In its guide to optimizing for generative AI features, Google makes clear that the basics of search engine optimization remain relevant for this. For Google Search, GEO is therefore not a replacement for SEO but builds on it.

Swipe sideways to see the whole table.

If your pages are hard to find in search, AI systems with web search will find them hard to reach as well. That is why GEO starts with clean search engine optimization.

AI systems answer questions from two sources: from what they learned in training, and from a current web search. For web search, the providers use their own crawlers. You control which ones you allow in robots.txt.

The most important crawlers at a glance

Swipe sideways to see the whole table.

There are also fetches triggered by individual users, such as ChatGPT-User and Perplexity-User. According to the providers, robots.txt rules do not always apply to these.

Example: allow search, block training

OpenAI shows in its documentation that each setting works independently. If you want to appear in ChatGPT search but not be used for training, it looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

According to their own statements, OpenAI and Perplexity pick up changes to robots.txt within about 24 hours. Whether you allow training is a business decision. For visibility in search answers, the search crawlers count.

1. Ask your target group

Write down which questions your customers ask an AI. For example: "Which tax advisor in Saarbrücken works with tradespeople?" Collect 10 to 20 such questions, from general to very specific. These questions are your starting point and later your measurement list.

2. Check what AI systems say about you today

Ask the questions from step 1 in several AI systems and record the answers. Note whether you are named, at which position and whether the details are correct. Repeat the test with the same questions so you can see changes. How to do this is covered further below under the prompt test.

3. Define your master data

Name, address, phone, services, founding year and contact persons. These details must read the same everywhere: on the website, in the Business Profile and in directories. Write them down once, binding, for example in a fact list, and take them from there for every new text.

4. Build a facts page

One page that clearly says who you are, what you offer, where you operate and what you do not offer. Short sentences, numbers with units and names instead of platitudes. We call such a page a grounding page. It is a normal, visible page of your website and just as readable for people as for machines.

5. Write answers that can be quoted

Every page should contain at least one sentence that holds true without context. Example: "AZ Onlinemarketing is an agency in Saarbrücken that manages Google Ads, SEO and AI visibility." AI systems can pick up such sentences. You find a before-and-after example in the next section.

6. Structure your pages

Use a clear heading order: one H1, then H2 and H3. Answer the question in the first two sentences of a section. Tables, lists and short paragraphs help people read. Google describes well-structured content with paragraphs and clear headings as a feature of helpful pages.

7. Add Schema.org data

Structured data based on Schema.org describes your organization, your services and your authors in machine-readable form. Make sure it matches the visible text. Google requires this in its guidelines and says at the same time: There is no special schema markup for AI features in Search.

8. Review your llms.txt

llms.txt is a text file in Markdown format that lists the most important pages of your website. Jeremy Howard proposed it in September 2024. It is not an official standard and guarantees nothing. Google writes that Google Search does not use such files, but that they do no harm either. It makes sense for other services and AI agents that read it. A separate guide on llms.txt will follow.

9. Align your details across the web

Directories, profiles and partner pages often show old details. Correct them where you can. Contradictions confuse AI systems: If an old phone number appears in three places, that very number can end up in an answer.

10. Maintain your reviews and your profile

A well-maintained Google Business Profile with real reviews provides trust signals. Google explicitly recommends keeping Business Profile details up to date, also with a view to the AI features of Search.

11. Clarify what you are not

A clear delimitation helps. If you do not offer web development, for example, say so on the page. That prevents wrong attributions, especially for names that exist more than once.

12. Measure at fixed intervals

Repeat step 2 at fixed intervals, for example monthly. Note what you changed. That way, you see what works. A single query is only a sample, because AI answers fluctuate.

AI systems like to pick up sentences that fully answer a question. The following example shows a made-up business so you can see the difference.

The second text names company, place, service, target group, area, a verifiable date and a delimitation. Every sentence holds true even if an AI quotes it on its own.

How to check your texts

• Does the company name appear in the sentence instead of "we"?
• Is the place named, instead of "in the region"?
• Is the service named concretely?
• Do numbers have a unit and a source?
• Does the sentence hold without the paragraph before it?

Measurement is the part many people skip. Without it, you do not know whether your changes do anything. Here is how:

1. Set the prompt list: Take 10 to 20 questions from step 1. Do not change them afterwards, or you cannot compare.
2. Choose systems: Test in the AI systems your target group uses, for example ChatGPT, Gemini, Perplexity and Google Search.
3. Ask neutrally: Use an account without history or a private mode, and do not name your company in the question.
4. Record the result: Enter every answer in a table.
5. Repeat: Test at fixed intervals and note what you changed in the meantime.

Template for your measurement table

Swipe sideways to see the whole table.

For Google Search, you have an additional data source. In its guide to AI optimization, Google recommends the generative AI performance report in Search Console. It shows how your content performs in the AI features of Search and Discover.

Many recommendations circulate around GEO. In May 2026, Google published its own guide, which puts several of them in perspective. For Google Search, this applies:

This applies to Google Search. Other providers such as OpenAI or Perplexity work with their own crawlers and their own rules. That is why the prompt test is worth running in several systems.

GEO does not replace good service or customer reviews. It makes sure that correct information about you is available. How often AI systems name you also depends on your other visibility, for example on SEO. In its own guide, Google advises caution with third-party tools that promise ranking success.

GEO (Generative Engine Optimization): Optimization of website and business details for answers from AI systems.

AEO (Answer Engine Optimization): Another name for the same goal, appearing in answers instead of only in result lists.

AI Overviews and AI Mode: Features of Google Search that generate answers with links to sources. In German, Google calls them "Übersicht mit KI" and "KI-Modus".

Crawler: A program that fetches web pages. Each AI provider has its own crawlers for search and training.

Grounding: An AI system bases its answer on current sources from the web instead of only on its training.

Entity: A clearly identifiable thing, such as a business with a name, place and services.

llms.txt: A proposed text file that lists the most important pages of a website for AI services. Not an official standard.

Frequently asked questions

What is GEO optimization?

GEO optimization means preparing your website and your details on the web so that AI systems such as ChatGPT, Gemini or Perplexity find your business and describe it correctly. The term stands for Generative Engine Optimization and comes from a research paper from 2023.

What is the difference between GEO and SEO?

SEO optimizes for rankings in search. GEO optimizes for AI systems naming you in their answers and describing you correctly. The two build on each other: Google names the SEO basics as decisive for its AI features too. You find an overview at our GEO agency page.

Do I need a dedicated facts page for this?

A clear facts page helps because it bundles all important details in one place. Whether it makes sense for you is shown by the check in step 2 of the checklist: If AI systems already describe you correctly, it is less urgent.

Should I block or allow GPTBot?

That is your decision. GPTBot collects content for training AI models. According to OpenAI, OAI-SearchBot is responsible for display in ChatGPT search. You control both separately in robots.txt.

Do I need an llms.txt for Google?

No. Google writes that Google Search does not use llms.txt and similar AI files and that they neither harm nor help. Other services and AI agents can read the file. It is therefore an addition, not a must.

How do I measure whether GEO works?

With a fixed prompt list that you test at fixed intervals in several AI systems. Record whether you are named and whether the details are correct. For Google Search, the generative AI performance report in Search Console adds to the measurement.

Which content do AI systems name more often?

Google describes helpful content as content that goes beyond common knowledge: your own experience, your own point of view, examples you created yourself, well organized with clear headings. A list of tips that anyone could write adds little new. For your business, that means: Show what you do in concrete terms, for whom and how.

How long does it take until GEO shows results?

There is no fixed time. It depends on when the crawlers fetch your pages again, how often an AI system uses web search and how much changes in your details on the web. According to their own statements, OpenAI and Perplexity pick up changes to robots.txt within about 24 hours. Everything else only repeated measurement shows you.

Can I do GEO myself?

Many steps you can do yourself: check master data, structure texts, align details. Measurement and technical implementation take more time than you expect. You get support through our AI visibility management.

Does GEO guarantee that I will be named in ChatGPT?

No. The providers decide for themselves what they name. GEO improves the foundations. There is no promise beyond that.