Is your brand invisible to LLMs? The guide to moving from SEO to GEO
As a consultant in Converteo’s media and acquisition practice, Quentin Barrat helps our clients improve their acquisition performance. A specialist in performance measurement and analysis, he implements customized solutions to understand their challenges and help them achieve their growth objectives.
The commercial web is changing. We decrypted the invisible infrastructure (UCP) in our first installment; today, we tackle the challenge of visibility. How do you exist when the interface disappears?
Key takeaways
- The end of the list: we are moving from a logic of choice (ten blue links) to a logic of response (a single summary). The AI no longer suggests; it asserts.
- The traffic paradox: being visible no longer means “being clicked.” Impressions are exploding, while incoming traffic is eroding. The goal is no longer to attract the user to your site, but to impose your brand in the AI’s response.
- The mechanics of entities: the algorithm doesn’t look for keywords; it seeks to connect concepts (knowledge graph). It simultaneously scans the press, technical databases, and reviews to verify the solidity of your brand “entity.”
- The mandate: GEO (generative engine optimization) consists of structuring your digital authority to maximize your citation score and your E-E-A-T (experience, expertise, authoritativeness, trustworthiness).
For two decades, the web’s unspoken pact remained unchanged: you provided content to Google, and Google paid you back in qualified traffic. This exchange model is collapsing.
With the arrival of generative answers (AI overviews, ChatGPT), we are experiencing a breach of contract. The user asks a question, the AI ingests your content, and it formulates the answer itself. The result: the user is satisfied, but they no longer visit your site. Studies already show that for certain informational queries, the click-through rate to the open web can drop by 20% to 40%.
Faced with this “great decoupling” between search volume (rising) and clicks (falling), panic is not the solution. Method is. SEO, the art of ranking, is giving way to GEO (generative engine optimization), the art of citation.
How can you ensure that when the AI speaks, it’s your brand it recommends?
GEO: from the “keyword box” to the “knowledge graph”
To master GEO, you have to unlearn classic SEO. A generative engine doesn’t work like a librarian who classifies books (ranking). It works like an investigator who connects evidence.
When you ask: “what is the best CRM solution for an SME?”, the AI doesn’t look for the page that repeats the word “CRM” fifteen times. It performs a vector semantic analysis:
- It breaks down the intent: it identifies the underlying concepts (cost, ease of use, integration), not just the words;
- It seeks consensus: it scans the web to see which brands are systematically and positively associated with these concepts by trusted third parties;
- It synthesizes: it writes an answer that presents the “consensual” solution as a truth.
A recent study by Princeton researchers on GEO showed that adding citations, statistics, and authoritative language could increase a brand’s visibility in AI responses by more than 30%. Your goal is no longer to be first on a list. Your goal is to become that statistical truth. In GEO, you don’t fight for a position; you fight for a mention.
4 levers to feed GEO
If the algorithm is an investigator, you must provide it with irrefutable evidence. For an AI to cite you, it must find corroborating signals across four distinct playing fields.
1. Authority and press: boost your E-E-A-T score
The AI is programmed to doubt what you say about yourself (commercial bias). However, it believes what Le Monde, TechCrunch, or a specialized journal says about you.
The lever: press relations are no longer a brand awareness tool, but a technical referencing tool. Every mention of your brand in a high-authority media (E-E-A-T) acts as a vote of confidence for the algorithm. This is the principle of “co-citation”: if you are cited alongside market leaders, the AI will associate you with that group.
2. Structured data: speak the machine’s language (schema.org)
Your website is no longer a visual showcase; it’s a data feed. The AI doesn’t “see” your design; it peels back your code.
The lever: if your product pages are not marked up in JSON-LD (schema.org/Product, schema.org/FAQPage), they are invisible. GEO optimization consists of making your content “liquid” and unambiguous for a robot. The more structured your data, the less effort the AI needs to understand it, and the more likely it is to use it.
3. Customer reviews and UGC: master social proof
To avoid recommending a bad product (the infamous “hallucinations”), the AI relies heavily on UGC (user-generated content). It analyzes the sentiment of reviews.
The lever: customer reviews and discussions on forums (Reddit) have become major ranking factors. A brand with no recent reviews or with negative sentiment will be systematically excluded from the final answer. It’s no longer enough to have a good rating (4/5); the review verbatim must contain the keywords the AI is looking for (e.g., “responsive customer service,” “fast delivery”).
4. Proprietary databases: feed Google Maps and Google Shopping directly
Google has its own maps: Maps and Shopping. To appear in Gemini’s responses, it is imperative to feed these databases directly.
The lever: the absolute completeness of your Google Business Profile and Merchant Center listings. This is the most easily accessible “food” for Google’s AI. Fill in every attribute, every optional field: it’s this level of detail that makes the difference between a citation and oblivion.
Writing for AI: fluency and density
Beyond the technical aspects, GEO changes the way we write. LLMs prefer content that is easy to digest and rich in facts.
- Fluency: favor simple sentences, a clear structure (H2, H3), and a direct answer to the question posed at the beginning of the paragraph. The AI hates filler (“fluff”).
- Information gain: to be cited, your content must provide unique added value (a new statistic, a new angle) compared to the existing consensus. If you repeat what others are saying, the AI has no reason to cite you as a source.
The blind spot: how to exist on ChatGPT?
A crucial nuance often escapes brands: not all AIs see the same web. While Gemini has access to the entire Google ecosystem, ChatGPT or Perplexity navigate blindly on certain closed data (like Maps).
To answer a local query (“plumber in Nantes”), these “alternative” agents fall back on often-neglected open sources: third-party directories (118000, Hoodspot) or well-structured store locators from brands. Paradoxically, being present on these “small” directories becomes a cutting-edge strategy to capture traffic from OpenAI.
From “share of voice” to “share of model”
We are entering an era where visibility is no longer measured in volume of clicks, but in share of model. The question is no longer “how many people saw my link?” but “how many times did the AI talk about me?”.
This transition offers a unique opportunity: to clean up your digital ecosystem. GEO doesn’t ask you to cheat the algorithm; it asks you to be a technically flawless brand, cited by third parties, and validated by your customers. It’s a return to the fundamentals of trust, but on the industrial scale of AI.
