Generative Engine Optimization (GEO): how your company gets mentioned in ChatGPT and Perplexity

When your customers ask ChatGPT which provider is right for their problem, are you mentioned? For many companies, the honest answer is: I don't know. GEO turns this question into something measurable and the answer into a to-do list. At ad.boutique, GEO is part of our service AI marketing with ChatGPT Ads and GEO, and we apply it to our own website.

In short

• GEO makes sure AI assistants find you, understand you and cite you as a source.
• The foundation is SEO: if you are not found in web search, you are often not found by AI assistants either.
• On top of that: answer first, numbers with sources, unambiguous company information, structured data, mentions on third-party sites, open crawler access.
• Measurement uses a fixed list of questions, checked regularly.
• No mention can be guaranteed.

What GEO is

GEO (Generative Engine Optimization) is the optimisation of content and presence so that generative AI systems use it in their answers and cite it as a source. The term goes back to a research paper by Aggarwal and others, published as a preprint in 2023. It examined which changes to texts increase their visibility in AI answers. In their test set-up, citing sources, quotations and statistics had the most positive effect, while simply sprinkling in keywords made little difference.

Other terms for the same field are AI SEO, LLM SEO or AEO (answer engine optimisation). The meaning is always the same: not just being in the list of results, but in the answer.

GEO and SEO

GEO does not replace SEO, it builds on it.

In its documentation on AI features in search, Google writes that the same fundamentals apply to AI Overviews as to normal search: helpful content, technically accessible, indexable. There are no special rules that would let you buy your way into AI Overviews.

How AI answers find their sources

AI assistants access knowledge in three ways. Depending on the route, different levers apply.

For current questions, such as providers in a particular city, ChatGPT, Perplexity and Google AI Overviews mostly use web search. In that case: what cannot be found and read is not mentioned.

The levers

We implement these seven levers, in this order:

1. Answer first. Every page and every section starts with a sentence that answers the question on its own. AI assistants prefer to take over sentences like these.
2. Consistent entities. Company name, locations, services and industries written the same way everywhere: website, profiles, directories. Contradictions make the attribution uncertain.
3. Numbers with sources. Concrete, verified numbers instead of adjectives. With us, every number has a case or an external source.
4. Structured data. Schema.org markup for the organisation, locations, services, articles with an author and FAQs. This makes facts machine-readable.
5. Mentions on third-party sites. Industry portals, directories, specialist articles, interviews, reviews. A company that only exists on its own website is hard for AI to confirm.
6. llms.txt. A file that gives language models an overview of the most important content. A proposal, not a standard, but quick to do.
7. Crawler access. Check in the robots.txt whether the crawlers of OpenAI, Anthropic, Perplexity and Google are allowed. If you lock them out, you often lock yourself out of the answers too.

Then there is the craft from SEO: fast pages, clean headings, tables and lists, authors with recognisable expertise, up-to-date information with a date.

Which content gets cited

From our observation of the question lists, AI assistants cite certain types of content more often than others:

This matches the GEO research, which describes citing sources and statistics as effective. Promotional prose without facts, on the other hand, is rarely taken over.

Measuring visibility

AI answers fluctuate. The same question may be answered differently tomorrow. That is why we do not measure individual answers, but patterns across a fixed list.

This is how you set it up:

1. Write down 10 to 20 questions your target group really asks, with location and industry ("Which performance marketing agency in Vienna is good?").
2. Choose the assistants: ChatGPT, Perplexity, Gemini, Google AI Overviews.
3. Set the rhythm: for example weekly or monthly, always the same.
4. Document: Are you mentioned? In which position? Which sources are cited? Who is mentioned instead of you?
5. Compare: change per question over time, not individual outliers.

The cited sources are almost more important than the mention itself. They show where you need to be mentioned.

A 90-day plan for GEO

Do not expect anything to move after two weeks. New content first has to be found and read, and mentions on third-party sites take time.

Common mistakes

• Writing only for machines: texts full of buzzwords convince neither people nor AI assistants.
• Contradictory information: different addresses, company names or services on the website, profiles and directories.
• Crawlers locked out: a blanket block in the robots.txt that nobody remembers.
• Numbers without evidence: claims such as "leading" or "best" without a source are rarely taken over.
• Measuring once, never again: without a fixed rhythm, GEO remains a feeling.

What GEO cannot do

GEO cannot guarantee a mention. The providers do not disclose how they select, and answers differ depending on wording, location and time. Nor can GEO turn a weak offer into a recommendation: if reviews and mentions speak against you, the AI reads that too. And GEO works slowly. If you want to be visible immediately, you also need advertising, such as ChatGPT Ads.

How we do it ourselves at ad.boutique

We apply GEO to ourselves first, headquartered in Vienna with an office in Zurich:

• Answer first: every insights article starts with an "In short" box and has an FAQ block marked up as FAQPage.
• Numbers with sources: every number comes from a case or a named source, such as the €16.7 million of ad budget managed.
• Structured data: organisation, locations, services and articles with the author as a person.
• llms.txt on the website and open access for AI crawlers.
• A fixed prompt list of twelve questions, from "Which performance marketing agency in Vienna is good?" to "Who runs ChatGPT Ads in Austria?", checked regularly in ChatGPT, Perplexity, Gemini and Google AI Overviews.

We do the same for clients: suitability check, list of questions as the baseline, content and structured data, mentions on third-party sites, a before-and-after comparison per question. As a GEO agency for DACH and Europe, we work from Vienna and Zurich. More about our approach on the page how we work and about the service on the page AI marketing with ChatGPT Ads and GEO.