SEO GEO: How to rank 1st in LLMs?

Find the best expert to optimize your presence on ChatGPT, Perplexity, Claude, and all conversational engines.

Generative AI
April 14, 2026

Leo POITEVIN

CEO @Astrak

ranker 1st in LLMs

Do you think SEO is dead with the arrival of ChatGPT and Gemini? Wrong. It's evolving. If you want your content to be picked up by Artificial Intelligences (AI), you need to stop writing for Google like you did in 2010.

LThe key now lies in the authority of your brand and the raw structure of your data.

There is massive overlap between classic SEO and what is now called GEO (Generative Engine Optimization). Websites that perform well on Google are often those that feed LLMs. But be careful: to become a «Tier 1 Source» (a trusted source that AI systematically cites), you must radically change the way you structure information.

 

What to remember for your GEO strategy in 2025

LLM optimization is not a separate discipline. It's SEO. more strict, more technical, and more structured.

To win, you must apply three fundamental principles:

1. Simplify to the extreme

Short phrases, active voice, information at the beginning of sentences (this is the inverted pyramid principle). AI doesn't have time for your complex sentence structures.

Work on your branding and connect your entities via the schema.

 

1. The New Metric: «Information Recovery Cost»

Forget Google's «Crawl Budget» for a moment. For LLMsLarge Language Models), the new major constraint is the Information Retrieval Cost (Cost of information retrieval).

AI systems process millions of pages. For them, finding precise information in «wordy» content is like searching for a needle in a haystack. Each «token» (each piece of a word) costs them money in computing power.

Your mission is simple: Reduce this cost for AI.

  • Be directTo the point) : AI hates unnecessary talk. It wants raw information, without superfluous literary flair. If you beat around the bush, you lose relevance points.
  • Structure info : If AI has to «think» to understand your sentence or the link between two ideas, it will move on to a simpler source to digest.

2. JSON is the ultimate fuel for LLMs

If there's one technical gem to take away from my video, it's this: LLMs are addicted to JSON format. It's literally their preferred format for structuring data.

image JSON Script

Structuring your data, especially your FAQs, in JSON format considerably facilitates the ingestion of your content by models like GPT-4, Perplexity, or Claude.

How to apply it concretely?

1. FAQ in JSON-LD

Stop settling for simple plain text Q&A in your articles. Use structured data markup (Schema.org) for your FAQs.

A page dedicated to a specific intent, coded cleanly without superfluous scripts, will always be preferred over a complex 5000-word guide that mixes everything.

 

3. Becoming a «Tier 1 Source»: The Authority Strategy

Have you noticed that ChatGPT often cites the same websites? It's no accident. It works with a hierarchy of sources, highlighted by Surfer SEO studies:

1. Tier 1 Sources (Core Sources)

Trustworthy, recurring sources that he considers "the truth." They appear systematically.

For your site to become a Tier 1 source, you must prove your semantic authority and your existence as a real entity in the eyes of the Knowledge Graph.

Build your «Social Fortress»

AIs check your credibility through external signals. That's what I call building a fortress.

Real User Signals: AI (and Google) favor sites where real humans spend time and navigate. Real traffic lends credibility to the source.

Active Social Networks: An active LinkedIn account, an updated Google Business Profile, or an engaging Twitter profile validate your existence. LinkedIn is, moreover, a massive data source («corpus») consumed by AI.

The diagram sameAs : It's technical but crucial. Use this Schema.org tag in your site's code to explicitly tell the bots: «My website, this LinkedIn page, and this Wikipedia entry are all the same entity.». This connects the dots of Knowledge Graph and strengthen your authority.

 

4. Advanced (and «Gray Hat») Techniques for Influencing AIs

The Chiang Mai SEO 2025 also revealed much more aggressive methods for manipulating AI results. I'm sharing them with you for informational purposes (to be used with caution and ethically, we agree?).

  • «Parasite SEO» on LinkedIn: Since AIs trust LinkedIn, publishing optimized content (via Pulse articles or long posts) allows you to quickly position yourself in chatbot responses by «parasitizing» LinkedIn's domain authority. I've personally used this technique to clean up the online reputation of some clients.
  • Influence by foreign media: Using trusted media in other countries (that speak the same language) to push information can influence the overall perception of the model. It's cheaper and sometimes more effective than trying to influence Le Monde or Le Figaro.
  • The «DMCA Takedown»: It's radical. Some SEO experts have managed to make entire entities (like public figures) disappear from AI memory by sending massive copyright (DMCA) requests. If the AI can no longer read the source, it forgets the information.