AI SEO agency: be the answer the AIs cite
Your customers no longer only search. They ask ChatGPT, Perplexity and Gemini. If the AI answers without citing you, you are invisible at the decisive moment. Our AI SEO agency combines the fundamentals of organic search with generative optimization: answer-first content, an entity graph, crawler management, and monthly citation tracking.
Cited, not merely indexed: monthly citation tracking across ChatGPT, Perplexity, Gemini and AI Overviews, robots.txt and llms.txt audited, on top of solid traditional SEO.
AI search optimization: why the moment is now
Search is fragmenting: Google and its AI Overviews, ChatGPT, Perplexity, Gemini, Copilot, Mistral. Six discovery channels, one question: when an AI answers on your market, is it your name that comes out? Classic search engines and answer engines now coexist on the same results page.
The end of the easy click
When an AI Overview appears, the click-through rate of the first organic position falls by 58% according to the Ahrefs study of February 2026. The answer is consumed without a click, the familiar zero-click pattern, and visibility shifts from the blue link to the cited source inside the generated answer. France moved to AI Overviews in the summer of 2026: our analysis of AI Overviews and AI Mode.
A channel that is expanding
AI traffic referred by generative engines has multiplied several times over in a matter of months across many sectors. It is qualified traffic: the user arrives after a reasoned answer in which your brand was recommended, which serves brand visibility directly and not only organic traffic. Generative share of voice becomes an indicator alongside Google positions, and we measure it: see our generative share of voice method.
Traditional SEO remains the entry ticket
No model cites a page it cannot read. Clean indexing, page speed, server-side JavaScript rendering for the pages that matter, coherent internal linking, a correct meta title on every page: that foundation conditions generative search as much as Google rankings. It also conditions the mobile experience, which models never see but your visitors do. A GEO agency that does not talk about technical SEO is selling half the work.
AI crawlers: what your site allows, and what it blocks without knowing
Almost no GEO agency treats this seriously, and it is the first place a citation strategy is won or lost. A site can do everything else right and stay invisible because a box was ticked in a security panel.
Training, Search, Agent: three uses, three decisions
Cloudflare now distinguishes three bot behaviours: Training, which collects content to train a model; Search, which indexes so it can answer questions about your content later; and Agent, which acts in real time on a person's behalf. Blocking Training protects your content from training. Blocking Search removes you from citations. They are neither the same bots nor the same decision, and most sites treat them as one block.
The Cloudflare change of 15 September 2026
From 15 September 2026, bots classified Training and Agent are blocked by default on ad-monetized pages of newly onboarded domains. Search bots remain allowed by default. The trap lies elsewhere: when a customer enables Training restrictions, multi-purpose crawlers such as Googlebot, Applebot and Bingbot fall with them, because the most restrictive rule applies. A box ticked on the AI side can cost you Google indexing.
robots.txt and llms.txt: what each one actually does
The robots.txt governs what crawlers may fetch: GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, CCBot. Each corresponds to a different use at the same publisher, and confusing them produces exactly the wrong result. The llms.txt, meanwhile, is not a standard recognised by the major engines. It is a useful convention, not a guarantee. We install it, without selling it as a magic lever.
Our GEO method: four pillars, one funnel
Google and the generative engines are not two worlds. They are two readers of the same site. The Contextual Density method from ProStar SEO serves both, in four pillars built in this order.
Technical foundations and JavaScript rendering
Core Web Vitals, indexing on Google and Bing, controlled crawl, and above all rendering: a page whose content only exists after JavaScript execution is readable by Googlebot in most cases and invisible to several AI crawlers, which do not render JavaScript. It is the quietest failure in generative search.
Answer-first content and citable formats
Pages restructured into direct answers, question-and-answer format, sourced and dated figures, comparison tables, marked-up FAQs. Models extract what can be cited in one sentence and ignore what requires a linear read of three paragraphs. Structured content answering an explicit search intent is what gets picked up. Structured, dated content also reduces the risk of hallucination about your brand, because it gives the model a verifiable version to reuse.
Entity graph and brand mentions
Consistency of your brand across the web: schema.org structured data, stable identifiers, press mentions, directories, profiles, marked-up author bylines, and a Wikidata attachment where it makes sense. Named entities and their relationships are what bring a brand into what a large language model treats as a safe answer. Backlinks still count, and unlinked mentions now count almost as much.
Citation tracking and generative share of voice
Monthly measurement across ChatGPT, Perplexity, Gemini, Copilot and AI Overviews, based on your real prompts rather than invented queries. The report says who is cited, on what, and which content gap explains the absence. The editorial backlog is built from that gap, and monthly reporting follows progress query by query against the cited competition.
AI engines and acronyms: the generative search landscape
Six engines, four acronyms, and a good deal of maintained confusion. Here is the decoder, then the detail on demand.
ChatGPT, Perplexity, Gemini, Copilot, Mistral, Grok
Each has its own index source, its own crawler and its own citation behaviour. Perplexity cites abundantly and links; ChatGPT, at OpenAI, cites more sparingly and depends on its search layer; Gemini leans on the Google index; Copilot on Bing. Mistral and Le Chat matter particularly for a French-speaking market, and DeepSeek and Grok remain marginal in Europe.
GEO, AEO, LLMO, GSO: what each acronym covers
GEO targets the citation inside a generated answer. AEO targets the direct answer, featured snippet included. LLMO refers to work on presence inside the models themselves. GSO is a commercial synonym for the first, and SGE was the former name of AI Overviews. The first three cover genuinely distinct realities: the detail is here.
AI Overviews and AI Mode: Google on both sides
Google is at once the search engine that ranks you and the answer engine that summarises you. AI Overviews appear above the classic results; AI Mode replaces the whole page with a conversation. Both feed on the organic index and on classic search results, which makes traditional SEO indispensable and insufficient. Opting out is possible, and rarely advisable.
The 6 AI engines where your brand needs to be cited
- Google AI Overviews: The generated answers at the top of Google's SERPs. When an AI Overview appears, the CTR of the first organic position drops by 58% (Ahrefs, Feb. 2026). Arriving in France with AI Mode announced for summer 2026 (neighboring-rights agreement of June 29).
- ChatGPT and ChatGPT Search: The most-used conversational engine. Its index relies notably on Bing: IndexNow indexing and answer-first content increase your chances of being cited.
- Perplexity: The AI engine that cites its sources most visibly. Highly responsive to sourced statistics and comparison tables.
- Google Gemini: Backed by the Google index and the Knowledge Graph: your entity graph (Wikipedia, Crunchbase, industry directories) directly influences your citations.
- Microsoft Copilot: Built into Windows, Edge and Microsoft 365: an underestimated B2B channel, fed by the Bing ecosystem.
- Claude: Used massively in enterprises. ClaudeBot must be allowed in your robots.txt and your llms.txt should give it a clear map of your content.
GEO, AEO, LLMO, GSO: the AI SEO acronym decoder
- GEO (Generative Engine Optimization): The dominant term in France: optimizing your content to be cited in AI-generated answers. It is the umbrella under which AEO and LLMO fit.
- AEO (Answer Engine Optimization): Historically the optimization of featured snippets, extended to AI Overviews: structuring each page to answer a question directly (answer-first format).
- LLMO (Large Language Model Optimization): The technical side: getting your brand into the models' knowledge (consistent entities, authority, llms.txt, AI crawlers allowed).
- GSO (Generative Search Optimization): A terminology variant pushed by some French agencies: it covers the same scope as GEO.
- AI SEO / AI search optimization: The mainstream phrasings, and what your customers actually type into Google. They all point to the same goal: being visible when AI answers in your place.
Our advice: do not choose an acronym, choose a method. All four cover the same work done well: citable content, consistent entities, measurement.
Technical checklist: preparing your site for AI crawlers
- robots.txt: allow GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended.
- llms.txt at the root: your editorial sitemap for language models.
- Structured data: FAQPage, HowTo, Organization, author with worksFor.
- IndexNow / Bing Webmaster Tools: ChatGPT Search relies on the Bing ecosystem.
- Answers at the top of the page: every key page opens with a direct 2-3 sentence answer (TL;DR).
- SSR or SSG: content served as static HTML, readable without executing JavaScript (this site runs on Astro SSG for precisely that reason).
12 signals that trigger a citation in ChatGPT, Perplexity and AI Overviews
- Direct answer up front: a 2-3 sentence TL;DR at the top of the page; AI engines extract first what immediately answers the query.
- Sourced statistics: dated figures with a source; ChatGPT and Perplexity favor citable, verifiable content.
- Comparison tables: the LLMs' preferred structured format for 'best', 'vs' and 'comparison' queries.
- FAQ marked up with FAQPage: question-answer pairs in Schema.org structured data, directly usable by AI Overviews.
- Entity authority: a brand that is consistent across Wikipedia, Crunchbase, LinkedIn and the press; the entity graph gets your name into the language models' memory.
- Marked-up author bylines: an identified author with worksFor markup; E-E-A-T weighs on source selection by Gemini and AI Overviews.
- AI crawlers allowed: GPTBot, ClaudeBot, PerplexityBot and Google-Extended in robots.txt; no access, no citation.
- llms.txt at the root: the site's editorial map for LLMs, a signal of openness to generative search.
- Careful Bing indexing: IndexNow and Bing Webmaster Tools; ChatGPT Search relies on the Bing ecosystem.
- Dated freshness: timestamped, updated content; Perplexity favors recent sources on newsy queries.
- High contextual density: more entities and relationships per page = more extractable material for the generative engines' RAG.
- Tracked generative share of voice: measuring your citations per AI engine and per query; what is not measured cannot be optimized.
Panorama: 15 AI SEO, GEO and AEO agencies active in France
The French AI search market is structuring fast: GEO pure players, AEO agencies born from the featured snippet, generalists adding an LLMO layer. Whether you are looking for a GEO agency in France, a partner to get cited by Perplexity or a complete generative search program, here are fifteen identifiable players, in no order of merit:
- Hack'celeration: An unapologetic AI SEO agency: 4 pillars (classic SEO, GEO content, entity authority, citation tracking) and a 'cited, not just ranked' pitch.
- AgenceSEOIA.fr: A pure player in AI search optimization, with SEO + GEO scoring and a visibility diagnosis across 6 channels (Google, AI Overviews, ChatGPT, Perplexity, Gemini, Copilot).
- SearchBooster / Oria: Champions the term GSO (Generative Search Optimization) in France and documents the differences between GEO, AEO and LLMO extensively.
- Oscar Référencement: A Paris agency that integrated a GEO offer and AI citation tracking early into its lead-oriented SEO.
- Upearly: Data-driven and GEO-ready: technical, semantic and popularity triptych strategies extended to generative engines and AI Overviews.
- SmartKeyword: Its Omnisearch covers Google, Amazon and generative AI (ChatGPT, Gemini), with a SaaS for managing visibility.
- Semji: An AI-augmented content marketing agency-platform, naturally positioned on answer-first content and marked-up FAQs.
- Optimize 360: Covers classic SEO and LLM optimization (llms.txt, structured data) for its healthcare, real estate and hospitality verticals.
- Première.Page: A dedicated AI agency offer, from generative search optimization and AEO to marketing automation.
- Agence Onze: A search studio with a structured AI SEO and ChatGPT visibility offer for SMBs, pragmatic and accessible.
- Natural-Net: A Bordeaux agency with a teaching bent, author of comprehensive GEO guides regularly cited by Perplexity and AI Overviews.
- Media-Dom: Documents the GEO, AEO, LLMO triptych and applies it to the AI search optimization of service websites.
- Eskimoz: The French search heavyweight has added generative search optimization and citation tracking to its full-search offer.
- Noiise: Its R&D division tracks the evolution of AI engines (ChatGPT Search, Gemini, Copilot) and equips its consultants accordingly.
- Yumens: A large multi-location agency gradually rolling out AI citation tracking and generative share of voice for its client networks.
Where do we fit? We claim no throne. Our angle: contextual density as the common language of SEO and GEO. The same content, measured 10% above the sources the AI engines already cite, serves both your Google positions and your generative share of voice.
| AI engine | Key citation signal | Priority GEO optimization |
|---|---|---|
| Google AI Overviews | E-E-A-T, structured data, SEO positions | FAQPage, direct answers, entity authority |
| ChatGPT Search | Bing index, freshness, citability | IndexNow, GPTBot allowed, sourced statistics |
| Perplexity | Recent sources, comparison tables | PerplexityBot allowed, dated answer-first content |
| Gemini | Knowledge Graph, Google entity graph | Schema.org Organization, entity consistency |
| Copilot | Bing ecosystem, B2B signals | Bing Webmaster Tools, marked-up expert content |
| Claude | ClaudeBot, llms.txt, clear structure | llms.txt at the root, static SSG HTML |
A terminology note: GEO agency France, AEO agency, LLMO agency, artificial intelligence SEO, generative search optimization, AI Overviews optimization... the market has not settled its vocabulary. In our ChatGPT visibility audits and citation tracking, we use "AI SEO" as the umbrella term: it is the one your customers actually type.
Contextual Density and GEO: why this method gets cited
Large language models reason in entities and relationships. A rich, structured page gives them more to extract. More reason, therefore, to cite you. We measure the contextual density and semantic diversity of the sources the AIs already cite on your queries, identify the entities and formats that trigger citation, and then produce optimized content that beats them on both axes. Content freshness weighs more here than in classic SEO: a model prefers a page dated this quarter.
The same asset serves your Google positions and your generative citations. That is the economic argument for GEO done properly: one editorial investment, two channels, one shared measurement. The approach applies to a regulated sector as readily as to an e-commerce site, and we use it on our own sector pages, from dental practices to law firms and plastic surgeons.
How much does an AI SEO agency cost?
The market is young and the pricing shows it: from a few hundred euros for a one-off audit to several thousand a month for a multi-market brand. The variable that matters is not the number of engines tracked, it is the volume of content to restructure and the state of the technical foundation you start from.
AI visibility audit
A one-off engagement: six engines queried on twenty prompts, a record of cited sources and their contextual density, an audit of allowed and blocked crawlers, a review of JavaScript rendering and structured data. Delivered with an SEO score, a GEO score, indicative pricing and a prioritised backlog you can hand to any other provider of GEO services or SEO services.
Monthly engagement
Restructuring pages into answer-first format, producing content against the identified gaps, work on the entity graph and brand mentions, link building, and monthly citation tracking. Monthly reporting documents each landing page created and its effect on citations. The common format for a single-market company.
Multi-market and multilingual brand
Several languages, different dominant engines by country, and an entity graph to keep coherent everywhere. The international side runs with our international SEO offer: models cite differently depending on the language of the question, even for the same brand.
FAQ: AI SEO and generative search optimization
What is the difference between SEO and GEO?
SEO targets positions in Google's results pages. GEO (Generative Engine Optimization) targets citations in the answers generated by ChatGPT, Perplexity, Gemini and AI Overviews. The foundations are shared (sound technicals, expert content, authority); GEO adds the answer-first format, the entity graph and citation tracking.
Is classic SEO dead now that AI Overviews exist?
No, it has become the entry ticket. AI engines rely on the indexes and signals of the classic web: a site that ranks poorly on Google has almost no chance of being cited by AI. Our approach handles both layers: SEO as the foundation, GEO as the extension.
How do you get cited by ChatGPT in France?
Four concrete levers: allow GPTBot and OAI-SearchBot in your robots.txt, take care of your Bing indexing (ChatGPT Search relies on it), structure your pages as direct answers with sourced statistics, and strengthen your entity graph (press mentions, directories, Wikipedia if eligible).
What is the llms.txt file?
An emerging standard file placed at the site root (like robots.txt) that presents your essential content to language models in a structured way. It is a signal of openness and a map of your site for AI. This site has one: agenceseoparis.com/llms.txt.
How do you measure visibility in AI engines?
With monthly citation tracking: we query ChatGPT, Perplexity, Gemini and AI Overviews on a panel of strategic queries for your market, and measure your generative share of voice against your competitors, query by query, engine by engine.
What budget for an AI SEO strategy in 2026?
French market offers range from about €3,000/month for an SMB to €25,000/month for large accounts. Our conviction: GEO should not be a separate line item but an extension of your SEO, which pools most of the cost. The AI visibility audit quantifies the gap before any commitment.
Does contextual density also help with AI?
It is actually its ideal playing field: LLMs reason in entities and relationships. A page whose contextual density exceeds competitors' by 10% gives the models more material to extract, and more reasons to cite you as a source.
AEO and featured snippets: how do they relate to AI Overviews?
AEO (Answer Engine Optimization) historically optimized Google's featured snippets: short answers, lists, tables in position zero. AI Overviews inherit that logic: answer-first content marked up with FAQPage and HowTo remains the most-reused format, by snippets and generative answers alike.
Does AI search optimization work for local SEO and regulated sectors?
Yes: ChatGPT and Gemini recommend local practitioners, firms and businesses by drawing on Google Business Profile, reviews and local entities. For YMYL sectors (health, legal, finance), E-E-A-T counts double: generative engines prioritize identifiable expert sources.
GPTBot, ClaudeBot, PerplexityBot: should you really allow them in robots.txt?
Yes, unless you have a strong editorial reason not to: blocking AI crawlers means giving up citations in ChatGPT, Claude and Perplexity. The right setup: GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended allowed, an llms.txt at the root, and content served as static HTML (SSG) that every AI engine can read without executing JavaScript.
AI Overviews are coming to France: what should you do first?
On June 29, 2026, Google announced the rollout of AI Overviews and AI Mode in France over the summer. Priorities: key pages restructured answer-first, E-E-A-T and entity graph reinforced, citation tracking active and generative share of voice measured from day one of the rollout, across ChatGPT, Perplexity and AI Overviews.
Generative search is built on the same foundations as organic search: our Paris SEO agency page sets out the full offer, sector pages such as construction SEO agency and CBD SEO agency show what that means in practice, and the about page introduces the SEO consultant behind the method.