Every AI-visibility checklist tells you to add llms.txt. Almost none tell you what to actually write in it.

I'll be straight with you: llms.txt is a nice-to-have, not a ranking factor. One shopping-assistant vendor described it, in their own documentation, as worth monitoring rather than prioritizing. The file won't make ChatGPT recommend your store. But it costs twenty minutes, it can never hurt, and it gives you control over how AI models describe your store when they do find you. That's worth doing — in the right order of operations, after the things that actually move the needle (product feeds, price consistency, review coverage).

Here's how to do it well.

What the file is

llms.txt is a proposed convention (a living spec, not a standard): a Markdown file at your site's root, yourdomain.com/llms.txt, written for language models rather than humans. When an AI assistant encounters your site, it may fetch this file the way a crawler fetches robots.txt — a shortcut to the information the model would otherwise have to piece together from scattered pages.

What to put in it (and the order)

1. H1: your store's name and one-line positioning. Not your tagline — what you actually sell, stated the way a customer would describe it. "Handmade ceramic dinnerware from Portland, Oregon" beats "Elevate your table experience."

2. A short background section. Two or three sentences: what you sell, who it's for, and one verifiable differentiator (e.g., "all glazes are lead-free and food-safe," "ships from warehouses in the US and EU"). Keep it factual. Models repeat what's in this file, and anything unverifiable becomes a hallucination with your name on it.

3. Product categories with links. A bulleted list linking to each category page, one line per category describing what it contains. This is the highest-value section: it tells the model how your catalog is organized.

4. Policies, one line each. Shipping, returns, and warranty — each with a link to the full page. Price filters are one of the most common shopping constraints ("under $100, ships here"), so your shipping policy link earns its place.

5. Optional: llms-full.txt. The convention supports a second, larger file with full page contents for the models that want it. For most stores under a few hundred products, the summary file is enough. Don't gold-plate this.

What NOT to put in it

  • Marketing fluff without facts behind it. The model will quote it, a customer will verify it, and you'll look dishonest.
  • Your entire catalog. This is a directory, not a database. Categories link out; the product pages carry the detail.
  • Prices. They change. Your product feed is the system of record for prices, and a stale price in llms.txt is worse than no price at all.
  • Anything you'd be uncomfortable with a competitor reading. It's a public file.

How to test it

Fetch yourdomain.com/llms.txt in a private window and confirm it returns Markdown, not a 404 or an HTML wrapper. Then run a quick sanity test: in ChatGPT, ask a question your store should answer using only your public pages, and check whether the answer tracks with what's in the file. You won't be able to isolate llms.txt's contribution precisely — models draw on many sources — but a mismatch tells you the file is being ignored or misread, which is actionable.

The honest priority order

If you have one afternoon: fix your product feed first, then price consistency between feed and page, then review coverage on third-party sites. Add llms.txt when those are done — it should take twenty minutes, and it's the kind of small polish that compounds. The stores that get recommended are the ones that got the fundamentals right and then did the small things anyway.

Want to see what else your store's AI-facing layer is missing? The free scanner checks robots.txt vs AI crawlers, llms.txt, product feeds, JSON-LD, and feed/page price consistency in about a minute: https://muse.ai/s/free-ai-visibility-checker-qxd5xhdxmnxbbww?utm_source=devto&utm_medium=social&utm_campaign=2026-10-merchant-tutorials