GEO: how to get cited by ChatGPT, Gemini and Perplexity
GEO, or generative engine optimization: how ChatGPT, Gemini and Perplexity pick their sources, what changes compared to SEO, and how to get your brand cited.
Published on November 20, 2026 · 6 min read
GEO (generative engine optimization), sometimes called AI search optimization, is the set of actions that help a brand be understood, accurately described and cited by AI assistants such as ChatGPT, Gemini, Claude or Perplexity. Where SEO aims for a good position in Google's results, GEO aims for a place in the answer itself. Both rest on the same foundations: a technically solid website and content that answers questions clearly.
In this article, we explain how AI assistants build their answers, what changes compared to SEO, the concrete levers for showing up in ChatGPT and its peers, how to measure your visibility and which mistakes to avoid.
What is GEO, in practice?
GEO means optimizing your brand's presence in the answers generated by conversational AI. In practice, the goal is that when a prospect asks an assistant "which provider should I choose for…" or "what is the best product for…", your company is part of the answer, and is described accurately.
The shift in behavior is real: some of your customers no longer just type keywords into a search engine. They ask full questions to an assistant, which replies with a synthesized text, sometimes alongside a few sources. If your brand is not in it, it simply does not exist for that prospect at that moment.
GEO therefore has three goals:
- Being cited: appearing in answers to the queries that matter for your business.
- Being described accurately: making sure the assistant presents your offers, products, prices or service area correctly.
- Being chosen: making the answer compelling enough that people click, get in touch or buy.
How do AI assistants build their answers?
AI assistants rely on two broad families of sources: what they learned during training, and what they look up on the web when you ask your question. The balance between the two varies by assistant and by mode.
Training data
A language model is trained on very large volumes of text. From it, the model builds a general representation of the world, of brands and of topics. That knowledge has a cutoff date: anything published afterwards is not part of it. A brand that is widely present and consistently described across the web is more likely to be "known" by the model.
Real-time search
More and more assistants query a search index to complement their answers with recent pages, which they then cite as sources. In that case, SEO mechanics become central again: a page that cannot be crawled, is not indexed or is poorly structured has little chance of being picked up.
Synthesis
The assistant then combines these elements to write an answer. It favors content that is clear, factual and easy to extract: a crisp definition, a list of steps, a direct answer to a question. This is where the format of your content makes the difference.
SEO vs GEO: what stays and what changes
GEO does not replace SEO, it extends it. Most search engine optimization best practices still apply, but the goal and the way you measure success evolve.
What stays the same:
- A fast, accessible website that crawlers can explore.
- Well-structured pages, with clear headings and a logical hierarchy.
- Useful, precise, up-to-date content that answers real questions.
- The importance of your brand's reputation across the web.
What changes:
- The goal: you are no longer only after a click from a results page, but after a mention in a written answer.
- Content format: direct answers, definitions and FAQs are more easily reused.
- Brand consistency: the assistant cross-checks several sources; contradictory information about your offer muddies its answer.
- Measurement: Google rankings are no longer enough, you need to look at what the assistants themselves say.
Concrete levers to show up in ChatGPT, Gemini and Perplexity
The most effective levers of AI search optimization are also the most concrete: content that answers, a readable structure and a consistently described brand. Here are the main ones.
- Answer questions directly: every important page opens with a clear answer in two or three sentences, then expands. It is the easiest format for an assistant to reuse.
- Structure your content: explicit headings, lists, numbered steps, short paragraphs. A well-organized page is easier to understand for humans and machines alike.
- Add FAQs: the real questions your customers ask, with short, precise answers.
- Use structured data: schema.org markup explicitly describes your organization, products, services, articles or FAQs. It makes your pages easier for search engines to understand.
- Create one page per offer: a specific page for each service or product, explaining what it does, for whom, where and how. Catch-all pages are hard to cite.
- Keep your brand consistent: same name, same description, same key information on your website, your profiles, directories, reviews and the articles that mention you.
- Keep your site technically flawless: load times, server-side rendering, indexing, clean markup. A framework like Next.js helps deliver pages that are fast and readable by crawlers.
- Consider an llms.txt file: a file placed at the root of your site that summarizes its content for language models.
The llms.txt file is an emerging convention, not a standard. Nothing currently guarantees that AI assistants read it. It costs almost nothing to add, but it never replaces clear content and a well-built website.
How do you measure your visibility in AI answers?
Measuring GEO means regularly asking AI assistants the queries that matter in your market and analyzing their answers. Without that measurement, there is no way to know whether your actions have any effect.
Three indicators are especially useful:
- Citation rate: across a representative set of questions, how often is your brand mentioned, and alongside which competitors?
- Answer accuracy: does the assistant correctly describe your offers, prices and features? Wrong information can cost you sales.
- Traffic from AI assistants: how many visits, contact requests or orders come from assistants, tracked in your analytics tools.
AI answers vary from one query to the next: a one-off check is not enough, you need to follow trends over time. That is the role of a tool like Skoup, an AI visibility platform for e-commerce built by Lunify: it measures what ChatGPT, Claude, Gemini and Perplexity say about a brand's products, compares their answers with the actual catalog and links visits from AI assistants to revenue.
Mistakes to avoid in AI search optimization
GEO attracts a lot of quick-fix promises. A few mistakes come up again and again and waste time.
- Dropping SEO: if your pages are not indexed or ranked, assistants that search the web will not find them.
- Mass-producing content: dozens of thin articles help neither readers nor AI. Fewer pages that are precise and useful work better.
- Looking for a miracle technical trick: no file or tag is enough to get cited. The fundamentals come first.
- Leaving contradictory information around: old prices, outdated addresses, inconsistent descriptions across different sites.
- Measuring nothing: without regular tracking of answers, you are optimizing blind.
- Talking about yourself instead of answering: purely promotional content is rarely reused. Content that answers the question is reused far more often.
GEO and AI search: the essentials
GEO is not a separate discipline, but the natural evolution of search optimization now that your customers ask their questions to AI assistants. It rests on a technically solid website, content that answers questions directly, a brand described consistently across the web and regular measurement of what AI assistants say. Companies that start early gain a head start, whatever their industry.
To go further, discover our SEO and GEO services, or let's talk about your visibility on Google and in AI assistants.
Frequently asked questions
Related articles
November 20, 2026 · 8 min read
Native vs hybrid app: how to choose, and when a web app is the better option
Native vs hybrid app, web app or PWA: user experience, offline mode, budget, app stores and SEO. The practical criteria for choosing the right mobile approach.
November 20, 2026 · 7 min read
How to build a SaaS: the key steps from idea to first paying customer
How to build a SaaS step by step: validate the problem, choose a business model, scope the MVP, set up a multi-tenant architecture, launch and iterate.
November 20, 2026 · 7 min read
Custom software or off-the-shelf SaaS: how to choose?
Custom software or off-the-shelf SaaS? Decision criteria, the hybrid approach, how a project runs, code ownership and maintenance: a guide to deciding.
A project in mind? Let’s talk about it.
Startup, SME or large group, whatever your industry: we help you frame the need and build the right product.