Being cited by ChatGPT is not luck and it is not a growth hack. It comes down to whether a retrieval system can fetch your page, understand who you are, and lift a clean statement from you without risk.
Here is the checklist, in the order we run it.
1. Confirm the crawlers are allowed in
Assistants use distinct agents: GPTBot and OAI-SearchBot for OpenAI, PerplexityBot for Perplexity, ClaudeBot for Anthropic, Google-Extended for Gemini's training access. Check your robots rules line by line. Many sites block these accidentally, inherited from a template or a security plugin.
Blocking is a legitimate choice for paywalled or proprietary content. It should be a decision you made, not a default nobody reviewed.
2. Check what the crawler actually sees
Fetch your own page without JavaScript. If the body is empty, an assistant may have stored a shell with your navigation and nothing else. Server-rendered or statically generated pages avoid the problem entirely, which is one reason we build client sites as static pages.
3. Give the model something quotable
Models paraphrase or quote short, self-contained statements. Pages that bury the answer in a narrative give them nothing safe to lift. Practical fixes:
- Open with a direct answer of 40 to 60 words.
- Use one clear H1 and a logical heading order.
- Keep paragraphs short and independently meaningful.
- Put comparisons in tables and processes in numbered steps.
- Include specific figures, names, dates and constraints.
4. Make your identity unambiguous
If your brand appears as three different names with two different descriptions across the web, you are an ambiguous entity — and ambiguous entities do not get recommended. Standardise the name, the category description, the services list, the locations and the people. Then mark it up with connected Organization, Service and Person schema, linking to every verified profile.
5. Show freshness and authorship
Publish visible dates and named authors with real credentials. A page that cannot prove when it was written or who wrote it is a weaker source than one that can, for both search engines and models.
6. Earn mentions where models already look
Assistants lean on third-party sources: review platforms, directories, comparison sites, community threads, industry publications. Being accurately described in those places matters more than another post on your own blog. Prioritise the sources you already see cited in answers about your category.
7. Cover the whole question chain
Buyers do not ask one question. They ask what something costs, whether an alternative is better, what goes wrong, and how long it takes. If you only answer the first question, you drop out of the conversation at the follow-up — which is usually where the decision happens.
8. Track it like a metric, not a vibe
Fix a prompt set. Re-run it monthly with identical wording. Record whether you appear, which competitors appear, which sources are cited, and how you are described. Watch the trend across runs; single answers vary because these systems are non-deterministic.
9. Correct misinformation quickly
Models sometimes state wrong facts about pricing, locations or services. The fix is not arguing with the model, it is correcting the underlying sources it is reading — your site, your profiles and the third-party pages describing you.
The honest expectation
Foundation work usually starts showing up in citations within four to eight weeks, with compounding movement once off-site corroboration lands. Nobody can promise a specific recommendation. What you can have is a number that improves and a reason for every move it makes.
Related reading: our AI Search Growth programme and LLM & ChatGPT SEO service explain the technical half of this in more detail.