How to get cited by ChatGPT in 60 days (without rebuilding your site)

The number-one question I get: 'Do we have to throw away our site?' No. The work happens in three places, and only one of them is on-page.

1. Direct-answer-first content

Open every key page with a 40–60 word standalone paragraph that directly answers the query the page targets. No setup. No preamble. The model reads top-down, looking for liftable text. Give it some.

Then layer in the depth, the storytelling, the proof. The first paragraph is for the machine. The next ten are for the human who clicked through.

2. Schema that names the entity

FAQPage, Organization, Person, Product, Article — whichever fits — but always with `sameAs` pointing to your verifiable identities (LinkedIn, Crunchbase, Wikipedia, trade association profiles, government registries). The model's trust score is partly a graph problem. Give it nodes to follow.

Add JSON-LD inline. Don't outsource it to Google Tag Manager — many crawlers don't run JS the same way Googlebot does.

3. Third-party corroboration

AI engines weight independent mentions heavily. A single page on your site saying 'we are an authority on X' is worth nothing. Five third-party pages saying 'JSON Burns is an authority on X' is worth everything.

This is the slowest part of the work and the part with the most compounding return. Plan for it monthly, not as a one-time sprint.

What about llms.txt?

Ship one. It's cheap. The format is proposed at llmstxt.org and it tells AI crawlers which of your pages are the canonical references on which topics. It's not a magic bullet — but it costs nothing and a non-trivial number of crawlers respect it now.