Insights · GEO

Get your brand cited
by ChatGPT.

A practical, no-hype checklist for Generative Engine Optimization — how AI assistants decide which brands to name, and the fixes that make yours one of them.

The short version: AI assistants don't rank pages — they compose answers, and they name the brands they can describe confidently. Getting cited is therefore a different job from getting ranked: it's about making your brand easy to identify, easy to verify, and easy to quote. Here's the checklist we run.

Why this is a different game from SEO

Classic SEO earns you a position in a list; the user still does the choosing. When someone asks ChatGPT, Gemini or Perplexity "who should I use for X?", the assistant does the choosing — it names a handful of options and describes them in a sentence each. If your brand isn't one the model can identify and describe with confidence, you're not ranked lower. You're absent.

The good news: the inputs are still largely under your control. AI engines learn about you from your own site, from structured data, and from what the rest of the web says about you. All three are workable.

How AI engines decide who to name

Different engines work differently — some answer from training data, some retrieve live pages, most blend both — but the brands that get cited tend to share the same traits:

  • A clear entity. The engine can tell exactly who you are: one canonical name, one domain, consistent descriptions everywhere you appear.
  • Quotable facts. Concrete, unambiguous statements ("a Hong Kong digital marketing agency, Google Partner, pricing from HK$2,500/mo") rather than slogans an engine can't safely repeat.
  • Machine-readable structure. Schema.org markup that states in JSON what your pages say in prose — organization, services, FAQs, prices.
  • Corroboration. Third-party pages that confirm what your site claims: directories, press, partner listings, reviews.
  • Crawlability. AI crawlers can actually fetch your pages — they're not blocked in robots.txt, hidden behind JavaScript-only rendering, or too slow to bother with.

The checklist

Work through these in order — the early items are the cheapest and move the most.

  1. Answer the buyer's question in plain prose, on a crawlable page. If the question is "best corporate caterer in Hong Kong", a page must exist that says, in so many words, what you do, for whom, where, and why you're credible. Engines quote sentences, not vibes.
  2. State your facts as facts. Name, location, category, credentials, pricing model, notable clients. Every fuzzy claim ("leading", "innovative") is a sentence an engine skips.
  3. Add Organization / Service / FAQPage schema. Structured data is the difference between an engine inferring what you are and being told. Keep it truthful and in lockstep with the visible page — mismatches erode trust in both.
  4. Publish an llms.txt. A plain-text summary of your site written for AI crawlers, at /llms.txt. It's a young convention, it costs an afternoon, and it hands engines a clean, canonical description of your business.
  5. Check your robots.txt. Decide deliberately which AI crawlers you allow. If you want citations, blocking every AI user-agent is self-defeating.
  6. Fix your entity consistency. The same name, the same one-line description, the same address and category across your site, LinkedIn, directories and partner pages. Contradictions make engines hedge — and hedging engines omit you.
  7. Answer the follow-up questions too. Real buyers ask "how much does it cost", "how long does it take", "is it worth it". FAQ sections that answer these plainly are among the most-quoted blocks on any site.
  8. Earn third-party mentions. An engine naming you takes a risk; corroboration lowers it. Partner directories, industry lists, press quotes and reviews all count — and they're checkable.
  9. Keep pages fast and server-rendered. Many AI crawlers read the raw HTML and move on. Content that only exists after client-side JavaScript may simply not exist for them.
  10. Test it, in the engines themselves. Ask ChatGPT, Gemini and Perplexity the questions your customers ask — in every language your market uses. Note who gets named, who gets cited, and what the engines believe about you. Then fix the gaps and re-test next month.

What "measuring GEO" actually means

There's no Search Console for AI answers yet, so measurement is manual but simple: a fixed set of buyer questions, asked monthly, in the engines your market uses, with the answers recorded. Presence, description accuracy, and who's named ahead of you. It's unglamorous — and it's the difference between a GEO program and GEO theater.

When to get help

Everything above is doable in-house with patience. Where an agency earns its fee is the same place it always has: doing it faster, across languages, with the technical pieces (schema, rendering, crawl policy) done right the first time. If you want a baseline before deciding anything, our free AI Readiness Score tells you where you stand — and it's yours whether or not you hire us.

Want your baseline first?

Find out if AI knows you exist.

Two free steps, in order. The AI Readiness Score reads your own site and rates the eight signals above out of 100, in about ten seconds. Then, if you want it, a person puts your buyer questions to ChatGPT, Gemini and Perplexity and writes up what they say — two working days, no obligation.