Your customers changed doors
For twenty years the question was "am I ranking well on Google?". It still stands, but a second one has settled in next to it: "when someone asks an AI to recommend a provider like me, do I exist in the answer?". ChatGPT counts hundreds of millions of weekly users, Perplexity is establishing itself as an answer engine, and Google itself now answers above its own results with AI Overviews.
The fundamental difference: a Google results page shows ten links, an AI answer cites two or three names. There is no page 2. Either you are in the answer, or you do not exist.
How AIs pick who to cite
The models draw on two sources: what they learned during training (masses of public text where your brand appears, or does not), and what they fetch live from the web at question time. On the first source you act slowly, through reputation. On the second you act fast: your website just has to be accessible, readable and trustworthy for their crawlers.
Why your brand is unknown to the models today, and where they get what they know: our article on why ChatGPT does not know your brand lays out the full diagnosis.
Project 1: let the AI crawlers in
First thing to check: do your robots.txt and your firewall let GPTBot, ClaudeBot, PerplexityBot and Google-Extended through? Many hosts and security plugins block them by default, with an understandable protective logic but a direct commercial effect: an AI that cannot read your website will never cite you on fresh questions.
The list of crawlers, what each one feeds and the exact lines to add: blocking or welcoming AI crawlers.
Project 2: be readable and citable
An AI crawler that enters your website needs to understand quickly: who you are, what you do, for whom, where. Two tools work for you:
- The
llms.txtfile, a plain-text summary at the root of your website that tells AIs what to read and what to remember. It is the little brother of robots.txt, aimed at comprehension rather than permission. The format, examples and pitfalls: the complete llms.txt guide. You can also generate yours in two minutes with our free generator. - Content that answers precise questions, with dated, quantified, attributed statements. AIs cite what they can extract cleanly: a page structured around sharp questions beats a vague presentation text.
The four conditions to meet, in order, to go from "readable" to "cited": GEO, how to get cited by ChatGPT and Perplexity.
Project 3: the reputation that builds trust
AIs cross-check. A business only cited by itself weighs less than one mentioned by a professional directory, the local press, customer reviews and partners. This groundwork is the same as classic SEO authority building. If that project is your priority, our guide on building recognition and earning Google's trust covers it in full; everything in it also serves your AI visibility.
What about Google in all this?
Google is no bystander: its AI Overviews answer directly at the top of the results, citing a handful of sources. The pages that earn their spot look a lot like the ones ChatGPT cites: a direct answer to the question, clear structure, identifiable source. Who loses traffic, who gains, and how to become the cited source rather than the casualty: AI Overviews in France, what changes for your visibility.
Measure yourself honestly
Three checks to run this month, in this order:
- Ask ChatGPT and Perplexity to recommend a provider in your trade and your city. Note who comes up, and whether you are in.
- Check access: robots.txt, firewall, llms.txt present and up to date.
- Repeat the test in three months, same questions, and compare.
Our free audit automates the measurable part: 23 checks dedicated to AI visibility (crawler access, llms.txt, structured data, citability), out of the audit's 107. You leave with a score and the exact list of what blocks you.
Rémy Ginoux
Romain Ben