How to get your site cited in ChatGPT and Google's AI answers

The click is leaking out of search

The mechanics of search are shifting under everyone at once. By several 2026 analyses, most Google searches now end without a click to an outside website — the answer is assembled on the results page or inside a chatbot before anyone leaves. Meanwhile, brands cited inside an AI answer earn meaningfully more of the clicks that do remain than they manage when the answer leaves them out.

So the goal quietly moves. Ranking first still matters, but the higher-value position is being the source an AI answer names. Chasing the vanishing click is a losing game; becoming the citation inside the answer is not.

There is no separate AI index to opt into

The most important thing to understand is that large language models do not conjure answers from nowhere — they extract from indexed, ranking pages. If your site is invisible in traditional search, it will be invisible in AI answers too, because there is no parallel AI index you can register for. Classic SEO is not dead; it is the foundation the whole thing stands on.

The practical baseline, then, is boring but non-negotiable: be crawlable, be indexed, and do not block the AI crawlers (OpenAI’s and others) in your robots rules. You cannot be quoted from a page nobody is allowed to read.

Write so a machine can quote you

Answer-engine optimization is mostly good structure taken seriously. Answer the question the page is about in its first sentence or two, before any windup — buried answers rarely get cited. Use descriptive headings, keep paragraphs focused on one idea, and use tables for comparisons and lists for steps, so a model can lift a clean, self-contained passage. Original data earns citations at a far higher rate than recycled roundups; if you can publish a number that originated with you, do.

Then make the structure legible to machines: Article, FAQ, and Organization schema, accurate metadata, and a semantic document outline. The signals that earn AI citations overlap heavily with the ones that earn strong rankings — you are not maintaining two strategies, you are hardening one.

How we build for it

We wire this in at the foundation rather than bolting it on. Every site we ship gets clean semantic markup, per-page structured data, prerendered content so crawlers and answer engines see finished pages instead of an empty shell, and an llms.txt map that hands AI agents a guide to the site. There is a satisfying overlap with our core discipline: the same qualities that make a page usable by a screen reader — real structure, honest headings, labeled content — are the ones that make it quotable by a model. Accessible and machine-readable turn out to be the same investment. For the flip side of this — protecting the rankings you already have through a rebuild — see how to redesign a website without losing your rankings.