In this article
More and more people ask their questions directly to ChatGPT, Perplexity, Gemini or Claude instead of going to Google. And when they do, they get an answer with two or three brands mentioned, not a results page. The question that matters for your brand is simple: are you in that answer, or not?
Showing up in AI-generated answers isn't luck. It's the result of a set of practices known as GEO. This guide brings together the checklist that gives your brand every chance of being cited, the mistakes that leave it out and how to measure the result.
What GEO is
GEO (Generative Engine Optimization) is the set of practices that makes your content eligible to be cited by AI assistants. It doesn't replace SEO: it adds to it. The full explanation is in what GEO is and how it works in practice.
The GEO checklist
1. Answer the question right at the start
Models extract direct statements more easily. Put the answer at the top of each section and expand afterwards. If the question is "how much does X cost", the first sentence should give the price or the price range. A short summary at the top of the article also helps.
2. Structure the content cleanly
Descriptive headings (H2 and H3), short paragraphs, numbered lists and comparison tables. The easier it is for a model to understand what each block does, the more likely it is to extract and cite the right information.
3. Be specific and verifiable
Avoid generalities like "innovative tailor-made solutions". Models value concrete facts: numbers, timelines, features, examples. When you use a number, say where it comes from.
4. Cite credible external sources
Content backed by studies, official data or recognised sources is easier to validate. Link to those sources instead of making unsupported claims.
5. Build authority on the topic
A model trusts sources it recognises as credible on a subject more. That comes from in-depth, consistent content around a topic, covering the main question and the doubts around it: the "how", the "how much" and the "why". It's better to be the reference in a niche than one more name in a broad topic.
6. Include structured data
Schema markup (Organization, Article, FAQPage), well-formatted FAQ sections and clear definitions help both Google and generative models interpret the content. FAQs are particularly useful, because they're already in question-and-answer format.
7. Let the assistants' crawlers in
Check that your robots.txt doesn't block the bots assistants use to read the web, such as GPTBot and OAI-SearchBot (OpenAI), ClaudeBot (Anthropic) and PerplexityBot. Some hosting and CDN services block these bots by default, so check the file your website actually serves, not just the one in your code.
8. Keep the content up to date
Fresh, consistent information earns trust. Visible dates, periodic reviews and consistency with what other reliable sources say reinforce your credibility. Outdated content loses weight.
9. Build a presence beyond your website
Assistants don't invent sources: they take them from places they already read and trust. Being present in relevant directories, industry publications and communities where your audience asks questions increases the surface from which you can be cited. The more reliable sites talk about you, the more the model finds you.
10. Test and measure
Take the key questions in your industry, type them into ChatGPT, Perplexity, Gemini and Claude, and record whether you're cited. Without measuring, you're optimising blind. The "How to measure" section below explains how.
An extra: the llms.txt file
llms.txt is a proposed convention, not a standard: a Markdown file at the root of the website with a summary of what the company does and links to its most important pages, meant to be read by language models. It doesn't replace robots.txt, which tells bots what they may or may not read. It's simple to create, but there's no guarantee that assistants use it or that it improves citation. Treat it as an extra, after the ten points above.
Mistakes to avoid
Many brands become invisible to AI because of concrete, easy-to-fix mistakes. These are the most common:
- Optimising only for keywords. Models work by meaning. Keyword-stuffed text with little substance is exactly what a model avoids citing.
- Burying the answer in the middle of the text. Three paragraphs of context before the point are bad for the reader and worse for a model, which extracts direct statements.
- Having no structured data. Without schema, FAQs or tables, the model has to guess what each block does, and the doubt works against you.
- Having no presence beyond the website. If your brand only exists on your own website, the surface from which you can be cited is minimal.
- Vague content with no facts. "Tailor-made solutions that boost your business" states nothing a model can cite.
- Letting content age. Old dates and statements that no longer match other sources make the model discard the page.
- Never measuring your presence in AI. Without testing, you don't know whether you're being cited or what to fix.
How to measure
GEO doesn't yet have a tool equivalent to Search Console, but you can already track it:
- Manual tests. Once a month, ask your key questions to the main assistants and record whether your brand appears, in what position and in what context.
- Tracking tools. There are paid tools, such as Profound or AthenaHQ, that automate these tests when the number of questions grows.
- Referral traffic. In GA4, track the visits coming from domains such as chatgpt.com and perplexity.ai.
GEO and SEO add up
GEO doesn't replace SEO. Almost everything that helps a page on Google (useful content, clean structure, technical quality, authority) also makes it citable by an assistant. The difference, and how to work on both together, is in GEO vs SEO.
Conclusion
Showing up in AI answers doesn't take magic; it takes method. Answer directly, structure well, be specific, let the assistants read your website, build a presence beyond it and measure the result. Brands that apply this checklist now will be present in the answers while their competitors are still debating whether it matters.
Frequently asked questions
What does it take to show up in ChatGPT?
Content that answers the questions in your industry directly and specifically, well structured, up to date and accessible to the assistants' bots, plus a presence in reliable external sources that talk about your brand.
How long does it take to show up in AI answers?
It depends on the industry, the competition and your starting point. Assistants that search the web in real time can reflect changes faster than those that rely only on what they learned in training. That's why it's important to measure every month.
Is llms.txt mandatory?
No. It's a proposed convention, with no guarantee that assistants use it. The points in the checklist carry more weight.
Does GEO replace SEO?
No. The two add up, and much of the SEO work also helps GEO.
Writes about AI consulting for operations and processes: how to map and simplify processes, apply AI where it has the most impact and measure the gains.