ChatGPT with its web browsing feature, Google AI Overviews, Perplexity, Bing Copilot: generative engines are multiplying and becoming a lasting part of search habits. For a premium brand that wants to appear in these responses, the question is no longer theoretical. It's practical: concretely, what should be done on a website to maximize visibility in these environments? The answer is less mysterious than it seems. LLMs value the same qualities as advanced traditional search engines but with particular emphasis on semantic clarity, information structure, and thematic depth. What sets them apart from a link-ranking algorithm is that they seek to understand, not just to index.
1. The seven levers of GEO optimization on a website
1. Semantic clarity of content. Every page must explicitly state what it is, who it's for, and what expertise it draws on. LLMs don't make complex inferences. What isn't clearly stated won't be understood.
2. Structured JSON-LD data. Implementing Schema.org is one of the most direct signals for AI engines. Organization, Service, FAQ, Article, Person, BreadcrumbList: every entity type must be described precisely.
3. Structured FAQs. FAQs are structures particularly well understood by LLMs, which look to answer questions. A well-written FAQ is a source models extract from preferentially.
4. Navigable tables of contents. A table of contents at the top of a long article helps LLMs understand the content's structure and extract the sections relevant to a given query.
5. SEO copy at the top and bottom of the page. This copy, which summarizes the page's positioning and expertise, is a primary signal LLMs use to qualify content.
6. Internal links between pages. Coherent internal linking helps LLMs understand the relationships between content and the site's thematic structure.
7. Precise meta titles and meta descriptions. These elements are used by LLMs performing web searches to qualify pages before reading them.
2. What Bonhomme implements in its projects
From its offices in Paris, the web agency Bonhomme systematically builds these seven levers into every website project it develops. Not as a checklist as a construction philosophy: every site must be understandable to a human exploring it for the first time, to a search engine indexing it, and to an LLM synthesizing it.
Concretely, this means: implementing precise meta titles and meta descriptions that exactly reflect each page's positioning and expertise; structured data JSON for key entities (organization, services, team, projects); structured FAQs on the most strategic pages; tables of contents on long-form content; SEO copy that explicitly states the agency's expertise; and internal linking that connects pages coherently with the overall positioning.
Beyond these technical elements, Bonhomme systematically recommends that its clients build a body of expert content Knowledge Center, blog, sector guides — that builds, over time, the thematic authority LLMs recognize and cite.
3. The specifics of ChatGPT and Google AI
When performing web searches, ChatGPT relies on Bing to access content in real time. It values content that directly answers precise questions, sources perceived as authoritative in their field, and clear structures that make information easy to extract. A site well optimized for Bing, and therefore for traditional SEO, is a good starting point.
Google AI Overviews, built directly into Google search results, draws on Google's index corpus. It values content that matches complex informational queries, sources with strong E-E-A-T authority (Experience, Expertise, Authoritativeness, Trustworthiness), and structures (FAQs, lists, definitions) that allow answers to be extracted directly.
In both cases, the best optimization stays the same: precise, structured, expert content, consistent over time.
4. Common recurring mistakes
• Publishing unstructured content. A long, continuous text with no headings, no lists, no FAQ, is hard for an LLM to read when it's trying to extract a precise answer.
• Ignoring JSON-LD. The absence of structured data is a significant missed opportunity for GEO visibility. It's one of the most direct and most underused levers.
• Optimizing only for keywords. Keyword-density logic belongs to SEO. GEO calls for thematic depth, covering a topic thoroughly, not just mentioning it.
• Neglecting long-form, expert content. LLMs prefer sources that cover a topic in depth over those that only skim it. A substantial 2,000-word article is worth more for GEO than ten short, superficial ones.
5. The Bonhomme vision
For the Bonhomme agency, optimizing for AI engines isn't a new, separate undertaking, it's the logical conclusion of a quality standard that has always guided its work. A well-built site, clearly structured, rich in precise and consistent content, is already a site optimized for LLMs. Most brands that have taken care of their online presence already have a head start without knowing it.
What the studio adds to this foundation is technical precision structured data, semantic linking, FAQs and a long-term content strategy that turns a good website into a source AI can cite.
"Optimizing for ChatGPT or Google AI starts with doing what we should always have been doing: writing clearly, structuring precisely, publishing regularly about what we truly know. AI just makes that standard more visible."
— Emmanuel Cruellas, co-founder & design director
"The best way to optimize a site for AI engines is to build it for demanding human readers. Content that satisfies an expert in their field will also satisfy an LLM."
— Morgane Urbain, co-founder & creative director