Module 05 · Strategy

GEO & New Search

⏱ 45 minutes
🟠 Advanced
📚 6 sections + quiz
🚧
Advanced module · intended for marketing, communications, and leadership roles

This module covers a technical and strategic topic: how to get referenced by AI search engines. It is particularly useful for marketing, communications, SEO, and decision-makers who own brand visibility. If you're new to AI and not directly concerned by search referencing, feel free to skip this module and move on to Module 6 on governance. You can always come back later if the need arises.

The silent earthquake

For 25 years, Google reigned over search. SEO was a proven science: you optimised your keywords, climbed the "10 blue links", and traffic followed.

Since 2024, this model has been cracking. Users no longer open 5 tabs, they ask their question to Perplexity, ChatGPT or Claude. Or they read the summarised answer directly in Google AI Mode. The verdict is clear: no click, no traffic.

📉
The number that hits

Some sites report up to -40 % of organic traffic since AI Overviews rolled out. Their SEO is identical. What changed is the mechanics of discovery.

Welcome to the era of GEO (Generative Engine Optimization).

The shift in user behaviour

Before: keyword search

  • Short queries: 1 to 2 keywords ("plumber Paris 11", "iPhone battery").
  • The user expects a list of links and navigates themselves.
  • Opens 3-5 tabs, compares, picks, leaves.

Now: conversational search

  • Long queries: 4+ words, often in natural language ("Which plumber is best for an urgent leak in the 11th on a Sunday evening?").
  • The user expects a direct, synthesised, sourced answer.
  • Asks follow-up questions in the same conversation.
  • Only clicks sources if the answer was convincing.
🔄
The objective has changed

SEO targeted the ranking (be number 1 on "plumber Paris"). GEO targets the citation (be the source ChatGPT uses to answer the question).

SEO vs GEO: a match across 4 dimensions

🔍 The classic SEO journey

1Short query ("plumber Paris")
2Google shows 10 blue links
3User opens 3-5 tabs
4Compares, picks
🎯Goal: rank on page 1

✨ The 2026 GEO journey

1Conversational prompt ("any plumber free tonight in the 11th?")
2Claude / Perplexity / AI Mode synthesises
3Direct answer + 3 cited sources
4User maybe clicks 1 source
🎯Goal: be cited as a source
Two parallel worlds today. Soon, the second one will dominate, and in some sectors it already does.
DimensionSEO (classic)GEO (new)
Objective Ranking in blue links Citation as a source in the AI answer
Queries Short isolated keywords Long conversational prompts
Authority Backlinks, domain authority Earned media: Reddit, forums, press, Wikipedia
Content format HTML optimised for Googlebot Markdown, structures readable by LLMs

"Machine-readable" content

LLMs love Markdown. It's their native training language. A well-structured Markdown page (hierarchical titles, lists, tables, FAQ) is vastly more cited than a poorly scanned PDF or a heavily visual site.

Four rules to make content "GEO-friendly"

  • Clear, hierarchical titles (H1 → H2 → H3, never skipped).
  • A direct answer at the start of each paragraph, followed by the justification (TLDR style).
  • Precise factual data: figures, dates, sources, the things that make your content "citeable".
  • FAQ sections with questions phrased as user prompts.

The llms.txt file

The new technical standard of GEO. The exact equivalent of robots.txt for Googlebot, except this one guides AI agents.

What it does

A well-structured llms.txt at the root of your site (yoursite.com/llms.txt) gives AI agents:

  • An overview of your site and what it covers.
  • Priority pages to index for different question types.
  • Zones you don't want used as a source.
  • Direct Markdown links to key content (instead of heavy HTML).

Example structure

📄 Example llms.txt file 
# Mantu · communications consulting agency

> Mantu is an international group (~12,000 employees) specialised
> in consulting, IT, transformation and communications.

## Main offerings
- [Strategic reputation consulting](https://mantu.com/...)
- [Employer brand and culture](https://mantu.com/...)
- [B2B marketing & influence](https://mantu.com/...)

## Case studies
- [Optic 2000 · Virtual Try-On](https://mantu.com/cases/optic-2000.md)
- [Fermob · Garden Styler](https://mantu.com/cases/fermob.md)
- [Pierre Fabre · Generative UX](https://mantu.com/cases/pierre-fabre.md)

## Methodology & expertise
- [Our Creative Intelligence approach](https://mantu.com/method.md)

## Optional
- [Careers](https://mantu.com/careers)
- [Legal](https://mantu.com/legal)
🎯
Measured effect

Sites that deploy a well-structured llms.txt report a significant rise in their inclusion rate in AI Overviews and in their citation rate by Perplexity / Claude / ChatGPT.

Off-site authority: the secret weapon of GEO

Here's the point that surprises everyone: AI models place enormous trust in external platforms, even more than in your own site.

Why

Reddit, specialised forums, Wikipedia, press, podcasts… these are places where humans freely discuss a brand, a product, an offer. For an LLM, that's a goldmine of "authentic signal", far more credible than your "Why choose us" page.

The top GEO-powerful platforms

  • Wikipedia: absolute reference. A well-crafted Wiki page shows up in 80 % of prompts about your topic.
  • Reddit: LLMs' favourite platform. Active presence in the right subs = massive citations.
  • Specialised forums: depends on the sector (Stack Overflow for dev, vertical communities elsewhere).
  • Trade press: expert articles in recognised vertical media.
  • YouTube: transcripts get parsed. Reference videos are quoted verbatim.
🎙
Concrete strategy

To position a brand in AI answers, editing your own site is no longer enough. You need to be present in the conversation: interviews in vertical press, active contributions in forums where your audience hangs out, expert podcasts. External brand-building becomes strategic again.

5-step GEO action plan

For a brand that wants to take the topic seriously:

  1. AI visibility audit. Ask Perplexity / ChatGPT / Claude "What do you know about [brand]?" across 20 varied prompts. Note what's cited, what's missing, what's wrong.
  2. Restructure your content in Markdown, especially "expertise", "case studies" and "methodology" pages. Clear hierarchy, FAQ, quantified data.
  3. Deploy a llms.txt at the root, with direct links to priority Markdown content.
  4. Activate your earned media: Wikipedia, Reddit, trade press, podcasts. External presence weighs as much as your own site.
  5. Measure and iterate. Redo the audit (step 1) every 3 months. Track the evolution of your citation rate.
🎯

Module 5 quiz

5 questions · 60 % to validate the module

1. What does GEO stand for?
GEO = Generative Engine Optimization. The set of techniques to ensure content is discovered, used and cited as a source by AI engines (ChatGPT, Perplexity, Claude, Google AI Mode).
2. What is the fundamental difference between the SEO and GEO objectives?
The objective has fundamentally changed. In SEO, you targeted ranking (top of the blue links). In GEO, you target citation as a source in the synthesised AI answer.
3. What is the llms.txt file used for?
llms.txt is the robots.txt equivalent for AI. At your site root, it gives AI agents a structured map of your content with Markdown links to key pages. Direct effect: rise in AI Overview inclusion rate.
4. What content format do LLMs love most?
Markdown is the LLMs' native training language: clear hierarchy, lists, tables. Well-structured Markdown content is massively more cited than heavy visual sites or poorly scanned PDFs.
5. Which external platform is particularly powerful for AI visibility?
LLMs place enormous trust in authentic external platforms: Reddit, vertical forums, Wikipedia, trade press. They read it as "human signal" and trust it more than a corporate page. External brand-building becomes central again.