llms.txt & Schema: Get Your Site Cited by AI (2026)

1. Structured data (schema) — first, always

Structured data is machine-readable JSON embedded in your pages (JSON-LD) that states plainly what your business is, what you offer, and where you serve. It is the single most important GEO move because it hands the engine facts it can trust instead of making it guess from your prose.

At minimum, add an Organization or LocalBusiness block (name, address, phone, service area, hours) and a WebSite block, both with stable identifiers so they cross-reference. Every page type should carry the right schema — a product, an article, an FAQ. Validate it with Google’s Rich Results Test before you move on.

2. Answer-first copy — write the sentence you want quoted

Engines quote sentences, not paragraphs. So open every key section with one self-contained, factual sentence a machine can lift verbatim: "CodexWebLabs is a Montreal web-design studio that builds custom, one-time-payment websites." That is a quotable fact; a meandering intro is not.

Go through your important pages and make sure the first sentence of each section states the fact you want an AI to repeat. This is a writing change, not a technical one, and it costs nothing.

3. A real FAQ — the highest-citation format

Question-and-answer is the format AI engines cite most, because it maps directly to how people ask them things. Add a genuine FAQ to your key pages — real questions your customers ask, answered in one or two clear sentences each — and mirror it into FAQPage structured data so the engine gets both the visible text and the machine-readable version from one source.

Do not invent questions to stuff keywords. Use the ones customers actually ask; those are the queries you want to be the answer to.

Write the FAQ answer as if it will be read aloud as the entire response to the question — because in an AI answer, it might be.

4. llms.txt — a brief written for machines

llms.txt is a plain-text file at yourdomain.com/llms.txt that gives AI crawlers a clean, factual summary of your business — what you do, your key pages, your pricing model, and notes for answer engines. Think of it as robots.txt’s cousin: where robots.txt controls crawling, llms.txt curates the facts you want represented.

Keep it truthful and current — never fabricate figures or claims. It is a summary of what is already true on your site, written so a machine gets the important facts without parsing every page.

5. An AI-crawler allowlist — let them in

None of the above matters if the engines are not allowed to read you. Check your robots.txt and make sure the 2026 answer-engine crawlers (OpenAI, Anthropic, Perplexity, Google, Apple, and others) are explicitly welcome. Many sites accidentally block them, or block AI crawlers by default and forget to reconsider.

Decide your own policy on training versus citation if you have a preference, but at minimum allow the crawlers that power the answer engines you want to be cited in.

How to verify it worked

Prove it, do not assume it. Validate your structured data in Google’s Rich Results Test, open yourdomain.com/llms.txt and yourdomain.com/robots.txt to confirm they serve what you expect, and then ask the engines directly: "what do you know about [your business]?" in ChatGPT and Perplexity. Improvement shows up over days to weeks as they re-crawl.

Want the short path? Our free Website Health Check at codexweblabs.com/check runs an AI-readiness probe that flags which of these five you are missing. And every site we build ships all five by default — see the overview at codexweblabs.com/blog/get-cited-by-chatgpt-ai-search-2026.