Being cited by an answer engine requires three things to succeed in sequence: the engine has to retrieve a page about you, extract a claim it can repeat with confidence, and verify that you are who the page says you are. Fail any one and you're absent from the answer regardless of how good the other two are.
Here are the five site-level things that have to be true for that sequence to work. This is the technical audit, not the strategy — the strategy is in the GEO field guide.
1. The content is in the HTML, not assembled by JavaScript
AI crawlers largely do not execute JavaScript. GPTBot, ClaudeBot, and PerplexityBot fetch your HTML and read what's in it — so if your prices, your claims, and your copy are rendered client-side, the crawler sees an empty shell where your best material should be.
This is the most consequential and least discussed item on the list, because it silently zeroes out everything else. View source on your most important page. If the substance isn't in the raw response, no amount of content work will help until it is. Server-render the pages that matter, or prerender them to static HTML.
2. You explicitly allow the AI crawlers
Check `robots.txt` for the specific agents — `GPTBot`, `OAI-SearchBot`, `ChatGPT-User`, `ClaudeBot`, `Claude-User`, `PerplexityBot`, `Perplexity-User`, `Google-Extended`. A wildcard allow usually covers them, but a lot of sites inherited a restrictive default from a security review or a CDN preset and never revisited it.
There is a real strategic choice here — some publishers block these agents deliberately to protect content. If you're a company that wants to be *recommended*, blocking them is unilateral disarmament. Decide it on purpose rather than inheriting it.
3. One page per question, with the answer in the first two sentences
Models extract; they don't read patiently. A page that spends four paragraphs building to its point loses to one that answers immediately and elaborates afterwards.
Concretely: one URL per real buying question, a heading phrased the way a person would ask it, and a self-contained answer in the opening lines that survives being lifted with no surrounding context. If the first sentence under a heading only makes sense after the preceding section, it can't be quoted — and quotable is the entire game.
4. Facts a model can defend
A model reaches for what it can stand behind. Named prices, dated figures, specific outcomes, cited sources. It avoids adjectives, because quoting *industry-leading* commits it to a claim it cannot support.
The Princeton GEO study found that adding citations, quotations, and statistics lifted visibility in generative-engine answers by up to 40% (Aggarwal et al., KDD 2024). Published pricing is the highest-leverage version of this in a vendor category, because almost nobody does it — which is precisely why the ones who do get quoted.
Structure helps extraction too: FAQ blocks where each answer stands alone, comparison tables, and schema markup (`Organization`, `Service`, `Offer`, `FAQPage`) that states plainly what you sell and what it costs.
5. One consistent entity across the web
The verification step. Your company name, address, and description have to match across your site, your schema, LinkedIn, and directory listings. A model that can't resolve you into a single confident entity will decline to name you — being unhelpful is safer for it than being wrong.
This is the one people skip because it's tedious and invisible from the inside, and it's frequently the actual blocker for a company that has done everything else well. A machine-readable statement of who you are — `Organization` schema, plus files like `llms.txt` and `ai.txt` that describe the site in plain terms — makes the resolution unambiguous.
The audit, in order
| Check | How to test it | Fails if |
|---|---|---|
| HTML-first content | View source on a key page | Your copy isn't in the raw response |
| Crawler access | Read robots.txt for named agents | A blanket disallow you inherited |
| Answer-shaped pages | Read the first two sentences under each heading | They need the preceding section to make sense |
| Quotable facts | Count the numbers on the page | Adjectives outnumber figures |
| Entity consistency | Compare name and address across five sources | Any two disagree |
Do them in that order. There's no point writing quotable answers into pages a crawler can't read, and no point being quotable if the engine can't confirm you exist.
What this doesn't do
Worth being honest: this checklist makes you *citable*. It does not, by itself, make you *cited*. Retrieval still depends on topical relevance and on third-party corroboration you don't control — which is why a company can pass all five and still lose to a competitor with a thinner site and a thicker footprint across the web. That gap, and how to tell which one you have, is covered in why your competitors show up and you don't.
The fastest way to find out where you actually stand is to measure rather than audit: the free citation audit runs your category's buying questions across the major assistants and reports where you appear and who appears instead. Fix the five above first, though — they're cheap, they're entirely within your control, and everything else is built on top of them.
