GEO: how to get your company cited by ChatGPT, Perplexity and Gemini

Buyers increasingly ask an assistant before they search: who can set up business development for a startup, what is a good alternative to a customer service BPO, which agency measures marketing on leads. The assistant answers in a paragraph and names a few companies. If yours is not one of them, the buyer may never reach a search results page.

Generative engine optimisation, GEO, is the work of making your company easy for those assistants to read, trust and cite. It builds on SEO rather than replacing it: an assistant that browses the web still needs pages it can fetch and understand. What changes is what it rewards.

How assistants decide what to say about you

Assistants answer from two sources: what they learned in training, and what they retrieve from the web when you ask. Training data is fixed and months old. Retrieval is where a small company can make a difference quickly.

When an assistant retrieves, it searches, reads a handful of pages and writes its answer from them. It favours pages it can read quickly, that answer the question directly, and that agree with what other sources say. It is cautious with claims it cannot confirm anywhere else.

Nine things to do, in this order

1. 01Ask first. Put the questions your buyers would ask to ChatGPT, Perplexity, Claude, Gemini and Google’s AI results, and write down what each says about you and your competitors. That is your baseline.
2. 02Make every page readable without JavaScript. Many crawlers do not run scripts; a site that only renders in the browser can look empty to them.
3. 03Let the crawlers in. Check that robots.txt allows the assistants’ crawlers, such as GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended, unless you have a reason to block them.
4. 04Publish one set of facts. Your name, what you do in one sentence, your services, markets and contacts should read the same on your site, LinkedIn, directories and partner pages. Assistants trust facts that agree.
5. 05Answer first. Open every service page and FAQ answer with the answer itself, in one or two sentences, then the detail. That is the part an assistant can quote.
6. 06Add structured data. Organization, Service, FAQPage and Article markup tell machines what a page is about without guessing.
7. 07Add an llms.txt. A short Markdown file at the root of the site that summarises the company and links to its key pages, ideally to clean Markdown copies of them. It is a proposed standard (llmstxt.org) and quick to add.
8. 08Earn mentions where assistants look. Directory profiles, client and partner pages, guest articles and press quotes are often what assistants draw on to describe a company. Your own site alone is rarely enough.
9. 09Check every month. Ask the same questions again, log the answers, and watch for visits from chatgpt.com, perplexity.ai and similar in your analytics.

What not to do

• Hidden text, or pages written only for machines. Search engines treat them as spam, and assistants that rely on search results inherit that judgement.
• Claims nobody else confirms. An assistant that cannot verify a number tends to leave it out, or report it as your claim.
• Separate tactics for every assistant. The groundwork is the same for all of them.

How this site does it

growaxis.net is built this way: every page is prerendered HTML, robots.txt welcomes AI crawlers, llms.txt links to a Markdown copy of every page, the full text sits in one llms-full.txt file, and the company facts match across the site and its profiles. It is the same groundwork GrowDigital installs for clients.

GEO will not make an assistant recommend a company it has no reason to trust. It makes sure that when there is a reason, the assistant can find it.