Your site is invisible to AIs. And it's a problem you haven't seen coming yet.

When one of your future customers asks a question in ChatGPT, Gemini or Perplexity, the AI answers in a few seconds with recommendations, company names, and precise references. In the vast majority of cases, your business is not listed. It's not a question of fame. It's a question of optimization, and most websites weren't designed for this new environment.

‍

A paradigm shift, not just another trend

Since 2022, the use of generative search engines has exploded. According to a Salesforce study published in 2024, 17% of professionals already use AI tools as the main source of information before making a B2B purchase decision. This figure rises to 41% among those under 40. Among 25-34 year olds, ChatGPT is already used as often as Google for professional searches.

It is not trivial. It is a structural shift in information behavior. And it's accelerating.

For years, the central question was: “Does my site appear on the first page of Google? ”. Tomorrow, it will be: “Is my business mentioned when an AI responds to my customers? ”. These two questions do not have the same answers, nor the same levers.

‍

SEO vs GEO: two logics, two languages

SEO (Search Engine Optimization) consists in optimizing a site to appear in Google results. It is based on hundreds of criteria: domain authority, keyword density, loading speed, backlinks. Google returns a list of links. The user chooses which to click on.

GEO (Generative Engine Optimization) works differently. Language models like Gpt-4o, Gemini 1.5, or Perplexity don't return a list of links. They generate a written response, built from thousands of sources that they have read, analyzed, and weighted. In this answer, they can mention brands, cite experts, recommend service providers. Or not mention anything at all.

The fundamental difference: SEO works on visibility in an index. GEO is working on credibility in the eyes of a model. They are two different targets, with different criteria.

‍

A concrete example

A purchasing manager is looking for a Webflow agency to redesign an institutional site. He types in ChatGPT: “Which specialized web agency Webflow should you choose in France for a B2B project?”

AI will synthesize the information available on the web and generate a structured response. She will mention agencies whose content she found clear, customer testimonials, well-written case studies, and blog posts on the subject. An agency whose site is well designed visually but lacking in structured content will probably not appear. An agency that writes regularly about its profession, with a clear positioning and detailed pages, has a good chance of being cited.

This is exactly where the battle for visibility in 2025 and beyond is being played out.

‍

What AIs are reading and why they cite some sources over others

Language models don't choose their sources at random. Their citation logic follows identifiable principles, which GEO practitioners have begun to map. Here are the four main signals.

‍

‍

1. The clarity and structure of the content

Language models prefer content that answers a question directly, with a logical structure: definition, explanation, example, conclusion. A page that starts with “We are a passionate creative agency” does not add any actionable information to a model. A page that explains exactly what the agency does, for what types of clients, with what results, is much more useful.

Concretely, this means: explicit page titles (“Webflow agency specializing in the redesign of B2B sites” rather than “We create digital experiences”), short and dense paragraphs, structured lists, clear definitions of technical terms.

‍

2. The thematic authority

A model assesses the credibility of a source by looking at whether it deals with its subject thoroughly and consistently. A company that publishes regularly on the same field, with varied and complementary angles, will be considered a reference on this subject.

This is called “semantic coverage”: does your site treat your domain comprehensively? Does it answer the questions your customers really have, not just the questions you want them to ask?

Example: a Webflow agency that publishes articles on the structure of CMS, best accessibility practices, the differences between Webflow and WordPress, the redesign process, and frequent mistakes will be much more likely to be cited than an agency whose site is limited to five static pages.

‍

3. External trust signals

AIs rely on what the rest of the web says about you. If media, industry blogs, specialized directories, customers, or partners mention your business and describe what you do, the model consoles this information and gives you greater credibility.

This is the GEO version of “netlinking” in SEO: not links to improve a domain score, but mentions to build a reputation readable by models. An interview in a sectoral media, a detailed customer testimonial published externally, a case study referenced by a partner: all this contributes to your AI share of voice.

‍

4. Technical markings adapted to LLMs

Beyond content, the way your site is technically structured plays an increasing role. Several standards are emerging:

• The llms.txt file: like robots.txt for search engines, this file tells AI model crawlers which pages on your site are the most important to read and how to interpret your content.
• schema.org markup: structured data that describes in machine language what you do, who you are, and what products or services you offer. A model that finds a well-filled schema Organization on your site has reliable and verifiable information.
• Structured FAQs: clear question-answer blocks, correctly marked up, are particularly well ingested by the models. They correspond exactly to the question and answer format that AIs are looking to generate.
• Full metadata: title, description, author, publication date, category. The more contextual and dated a piece of content is, the more reliable it is considered to be.

‍

Why act now and not in two years

The question we often hear: “We'll see in a few years if this trend really takes hold.” It's a strategic error, for two reasons.

‍

The use is already there, and it is growing rapidly

As of January 2025, ChatGPT had over 300 million active weekly users, up from 100 million two years earlier. Perplexity processes over 1 billion requests per month. Google itself has integrated its “AI Overviews” into its search results: a generative summary now appears at the top of the page for millions of queries, even before the classic links.

This shift is not going to be reversed. It's going to get bigger. AI tools are becoming more accurate, more reliable, and better integrated into business workflows. The search behavior of your customers is already changing.

‍

‍

AI visibility is built over time

Like SEO, GEO is not an immediate result. It depends on the quality and volume of content published, the consistency of the positioning over time, and the mentions accumulated on other sources. A business that starts today is building a lead. A company that waits in two years will have to catch up with two years of content published by its competitors.

To give an idea of magnitude: in SEO, it takes an average of 6 to 12 months for well-optimized content to start generating significant organic traffic. In GEO, signals accumulate in the same way. Each article published, each mention obtained, each page structured correctly contributes to building the representation that the models have of your business.

‍

What Norry is offering in practice

We have designed a GEO offer in three steps, applicable regardless of your sector and the maturity of your current digital presence.

‍

1. The AI visibility audit

The first step is diagnosis. We test how your brand is perceived by the main generative models in circulation: ChatGPT (Gpt-4o), Gemini, Perplexity, and depending on the sector, Claude or Mistral.

Concretely, we formulate the questions that your customers really ask these tools, we observe if your business is mentioned, how it is described, if the information is correct, and what is being said about your competitors. We measure your “AI Share of Voice”: the proportion of responses in which you appear on a set of target requests.

This audit results in a clear roadmap: which pages to optimize first of all, which pages to optimize, what content to create, what technical signals to put in place, what external mentions to look for.

‍

2. Optimizing content and structure

Once the diagnosis is made, we take action on your site. This covers several levels:

• Rewriting key pages: home page, “About” page, service pages. Each page should clearly answer at least one question that your customer asks before choosing a provider.
• The creation of content with high semantic value: blog articles, practical guides, case studies. This content should cover the angles from which your expertise can be recognized by a model.
• Technical markup: setting up the llms.txt file, enriching schema.org data, structuring FAQs, completing metadata.
• Work on external signals: identification of opportunities for mentions (directories, media, partners), support in the production of content cited by other sources.

‍

3. Monthly follow-up

AI visibility evolves based on model updates, new content published by your competitors, and your own production. We provide monthly follow-up that includes:

• A report from your AI Share of Voice on target requests compared to the previous month.
• An analysis of new content opportunities identified based on the evolution of generative responses.
• Recommendations for technical and editorial adjustments.
• A monthly strategic point to align our actions with your business goals.

‍

The question is no longer whether to do it, but when.

GEO is not just another trend. It is a fundamental change in the way information flows and in the way decisions are made. Businesses that embraced SEO before others built sustainable positions for twenty years. There is no reason why this cycle should not happen again with GEO.

Your site may be excellent: well-designed, well-written, well-referenced on Google. But the question that matters now is different: when your future customer asks an AI question about your sector, does your name appear in the answer?

If you don't know it, that's already an initial response.

‍