ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews no longer just rank pages: they retrieve passages, cross-check them and write an answer. We make sure yours is part of it, and gets cited.
Visualisation: an embedding space. The query (white dot) retrieves its nearest neighbours. The goal of GEO: that these are your contents (in red).
GEO (Generative Engine Optimization) covers the techniques that increase the likelihood that a generative engine retrieves your content, understands your business and cites it in its answer: ChatGPT, Gemini, Claude, Perplexity, Copilot, as well as Google's AI Overviews and AI Mode.
SEO puts you in the list of links; GEO gets you into the answer. The two complement each other, because generative engines draw heavily on search indexes.
At JSANCHEZ, the AI visibility audit is free and GEO packages start at €390 excl. VAT per month (see pricing).
Between your customer's question and the answer displayed, five steps decide who gets cited. GEO acts on each of them.
› prompt "GEO agency for an SME in Lyon?" › fan-out [GEO pricing SME] [GEO agency reviews] [GEO vs SEO] [AI visibility audit] › retrieve 120 passages → top-k = 12 › rerank reliability · freshness · consensus › answer 3 sources cited [1] [2] [3]
prompt
“Which agency can make us visible in ChatGPT?” The question is interpreted with its context: language, location, conversation history.
map your customers' real prompts
query fan-out
The engine breaks the question down into sub-queries run in parallel: prices, reviews, comparisons, alternatives, proximity.
cover every related intent
retrieval
Search in a web index (Google, Bing or proprietary) and by vector similarity: only the closest passages (top-k) are kept.
self-contained passages, readable in HTML
rerank + grounding
Passages are re-ranked according to the source's reliability, freshness and corroboration between independent sources.
consistent entity, third-party evidence, dates
generation
The LLM writes a summary and cites a handful of sources [1] [2] [3]. Being selected here means being recommended.
quotable answer sentences, named brand
No magic recipe: information engineering, applied to your brand.
Your business must exist as an unambiguous entity: a schema.org JSON-LD graph linked by @id, sameAs properties pointing to your official profiles, Wikidata where relevant, identical name, address and phone number everywhere, and a complete Google Business Profile.
Passage-level optimisation: each section answers in one or two sentences, then goes into detail. One passage = one idea, understandable once extracted. Question headings, definitions, tables, dated and sourced figures.
We map the sub-queries that engines generate around your services (prices, reviews, comparisons, alternatives, proximity) and cover each intent with a dedicated page or passage.
robots.txt configured user agent by user agent, server-side HTML rendering (OpenAI, Anthropic and Perplexity crawlers do not execute JavaScript), sitemap with lastmod, IndexNow, llms.txt file, server response time.
Models cross-check: the more consistently your brand is described by third-party sources (directories, press, comparison sites, reviews, forums), the more often it is selected. We work on co-citations and get incorrect information corrected.
A panel of prompts run every month on each engine, several times, because answers vary: citation rate, AI share of voice, position in the answer, accuracy. On the website side: traffic from chatgpt.com, perplexity.ai or gemini.google.com and AI crawler logs.
Being a pioneer has to be proven. This website is our lab: everything we offer is live here, and anyone can check it.
@id, with company registration and VAT numbers to remove any ambiguity about who we are.$ curl -s https://www.jsanchez.fr/llms.txt | head -3 # Agence JSANCHEZ > Agence de référencement fondée en 2016 sur la Côte d'Azur… $ curl -s https://www.jsanchez.fr/robots.txt | grep -c "User-agent" 20 $ # excerpt of the JSON-LD graph on every page { "@type": ["ProfessionalService", "Organization"], "@id": "https://www.jsanchez.fr/#agence", "foundingDate": "2016-05-09", "knowsAbout": ["Generative Engine Optimization (GEO)", "Retrieval-Augmented Generation (RAG)", "Query fan-out", …] }
| SEO (search engine optimisation) | GEO (generative engines) | |
|---|---|---|
| Unit optimised | The page (URL) | The passage and the entity |
| Mechanism | Inverted index and ranking algorithms | Retrieval (index + vector similarity), reranking, generation by an LLM |
| Key signals | Relevance, links, technical health, Google Business Profile, reviews | Extractable passages, entity consistency, structured data, co-citations, freshness |
| Visible result | List of links, Local Pack | A written answer citing a few sources |
| Where | Google, Bing | ChatGPT, Gemini, Claude, Perplexity, Copilot, Google's AI Overviews and AI Mode |
| KPIs | Rankings, CTR, clicks, conversions | Citation rate, AI share of voice, accuracy, traffic from AI |
A panel of real prompts run on 5 engines, several times each: who is cited, from which sources, with which errors. This is your dated starting point.
schema.org graph, sameAs, NAP consistency, Google Business Profile, robots.txt, HTML rendering, llms.txt and IndexNow.
Pages and passages covering each sub-query, answer first, dated and sourced figures, comparison tables.
Co-citations, reviews, third-party mentions. Monthly report: citation rate per engine, accuracy, traffic from AI.
Monthly prices excluding VAT. The AI visibility baseline is free.
Twelve concepts to understand how an AI chooses its sources.
sameAs).GEO delivers its best results when backed by solid SEO, Google Ads campaigns for urgent searches and a fast, well-structured website. At JSANCHEZ, all three are run by the same person.
We ask ChatGPT, Gemini, Claude, Perplexity and Google your customers' questions, and show you who is cited instead of you.