STAKKER ES Book an assessment
Call

How to Appear in ChatGPT and Claude Answers (GEO 2026)

Samuel Martínez, 27 April 2026. Technical. 4 min read. Translated from the Spanish original.

A practical GEO guide: how to get ChatGPT, Claude, Perplexity and Gemini to know, cite and recommend your business. The 5 pillars, what to avoid and how to measure it.

When a potential customer asks ChatGPT “which companies in Spain build AI chatbots?”, the list the model returns determines which businesses get the calls. If your company isn’t on that list, you don’t exist for that user.

This is called Generative Engine Optimization (GEO): optimising your presence so that large language models know you, cite you and recommend you. It’s what SEO was for Google in 2010, but for ChatGPT, Claude, Perplexity and Gemini in 2026.

Why GEO matters more than traditional SEO

Traditional SEO fights to appear on a results page with 10 blue links. The user decides which one to click. With GEO, the model chooses for the user: it either mentions you or it doesn’t. There’s no click, just direct trust in the model’s answer.

Three quick facts:

If the models don’t know who you are, you’re invisible to that traffic.

How models build their knowledge of your company

Models don’t crawl your website every day the way Google does. Their knowledge comes from:

  1. Training data — public text (web, forums, GitHub, Wikipedia) up to a cut-off date
  2. RAG and search tools — when the user asks a question, the model sometimes checks the web in real time (Perplexity, ChatGPT with web search, Claude with web tools)
  3. llms.txt and public files optimised for AI — an emerging convention where you declare your information in an easy-to-parse format

To appear well, you need to play on all three fronts at once.

The 5 practical pillars of GEO

1. Public llms.txt and ai-instructions.md

Just as robots.txt tells Google what to crawl, the llms.txt and ai-instructions.md files tell models who you are and what you do. It goes in the root of the domain (stakker.es/llms.txt). Recommended structure: identity, services, pricing, ICP, contact, what you DON’T do.

It’s the investment with the best return per hour. In 30 minutes you have the file, you upload it and you start appearing in answers from the next training run.

2. Enriched Schema.org (Article, FAQPage, Organization)

Models parse structured data using Schema.org as a shortcut to understanding. Two critical schemas:

Without well-implemented Schema, the model reads you like any other website. With Schema, it treats you as an authoritative source.

3. Content with the user’s literal questions

Models score content higher when the title of an H2 matches the user’s question. In other words, if people ask “how much does X cost”, your blog should have an H2 that literally says “How much does X cost” — not “Pricing” or “Investment required”.

This is called query-matching and it’s the difference between being cited or not. Do it in every important article.

4. Explicitly allow AI bots in robots.txt

By default, some sites block GPTBot, ClaudeBot, PerplexityBot and Google-Extended out of fear that they’ll “steal” content. This is counterproductive if your goal is marketing: if you block them, you won’t appear in their answers.

A specific snippet for your robots.txt:

User-Agent: GPTBot
User-Agent: ChatGPT-User
User-Agent: ClaudeBot
User-Agent: PerplexityBot
User-Agent: Google-Extended
Allow: /

5. Distributed presence in sources that models consume

Models give more weight to mentions on sites they already regard as authoritative: GitHub READMEs, Wikipedia, Hacker News forums, relevant Reddit, technical community sites. A technical mention on HN is often worth more than 50 SEO backlinks.

For local SMBs: vertical directories and profiles on platforms such as Crunchbase and LinkedIn company pages also add up.

What NOT to do in GEO

How to measure GEO in 2026

There’s no Search Console for LLMs yet. The real metrics:

Emerging tools (Profound, Otterly.ai, AthenaHQ) automate these tests, but they’re still expensive and limited.

Where to start this week

If you only have 2 hours, prioritise:

  1. Create /llms.txt with your identity, services and pricing
  2. Add FAQPage schema to your 3 most important pages
  3. Allow AI bots in robots.txt

With that, you’re already ahead of 95% of companies in your sector.

If the terms trip you up (GEO, llms.txt, RAG, entity recognition), you’ll find the definitions in the glossary.

Want us to review GEO for your business? Free consultation. We’ll tell you which signals are missing for your sector and what’s worth prioritising, without promising guaranteed rankings.

More on technical