llms.txt for Shopify stores: how and why

llms.txt is a plain-text Markdown file at the root of your domain that hands AI assistants a curated map of your most important pages — collections, hero products, policies, guides — each with a one-line description. It does not replace SEO or schema. It gives models a clean, human-written index instead of leaving them to crawl and guess.

I am a senior Shopify developer in Delhi, and I build and tune stores for D2C brands in India, the US, UK, and UAE. I have added llms.txt to production stores and watched what it does — and does not — change. This is the honest version: what it is, whether it is worth your time, how to wire it onto Shopify given the platform's quirks, and exactly what to put inside.

What is llms.txt, in plain terms?

llms.txt is a proposed standard for a file you publish at yourstore.com/llms.txt. It is written in Markdown, and it does one job: it points AI assistants at the handful of pages on your site that actually matter, with a short description of each.

The problem it solves is real. When ChatGPT, Perplexity, Claude, or Google's AI answers need to understand your store, they either crawl your live HTML — which is heavy with navigation, scripts, popups, and tracking — or they rely on whatever they already indexed. Your actual signal (what you sell, who for, your best pages) is buried in markup that was built for browsers, not readers. llms.txt flips that. It is a short, opinionated document written for a model that reads Markdown, telling it exactly where the good content lives.

A minimal example looks like this:

# Aurelia Skincare

> Clean, dermatologist-tested skincare for Indian skin and climate. Direct-to-consumer, ships across India.

## Bestsellers
- [Vitamin C Serum](https://aurelia.com/products/vitamin-c-serum): Our top seller — brightening serum for pigmentation and dull skin.
- [Barrier Repair Moisturiser](https://aurelia.com/products/barrier-moisturiser): Fragrance-free daily moisturiser for sensitive skin.

## Collections
- [Full range](https://aurelia.com/collections/all): Every product, by concern and skin type.

## Policies & Guides
- [Shipping & Returns](https://aurelia.com/policies/shipping): Free shipping over ₹999, 15-day returns.
- [How to layer actives](https://aurelia.com/blogs/guides/layering): Which serums to combine and which to avoid.

## About
- [Our story](https://aurelia.com/pages/about): Who we are, what we test, where we make it.
- [Contact](https://aurelia.com/pages/contact): Support hours, WhatsApp, email.

That is the whole idea. A title, a one-line summary, and grouped links with descriptions. It is deliberately small — a curated table of contents, not a database dump.

Why would this help my store get cited by AI?

Here is where I will be straight with you, because a lot of the marketing around llms.txt oversells it.

As of 2026, no major AI assistant has publicly confirmed that it reads llms.txt as a ranking or citation signal. Some tools and crawlers in the ecosystem do look for it; the big consumer assistants mostly stay quiet. So the honest framing is not "publish this and get cited" — it is "publish this because it is cheap, it can only help, and the process of writing it makes everything else better."

Three concrete reasons it is worth doing anyway:

1. It is a forcing function for clean answers. To write good one-line descriptions, you have to decide what each page is actually for, in plain language a customer would recognise. That clarity is exactly what AI assistants extract and quote. The file is almost a side effect; the thinking is the value.
2. It reduces ambiguity for tools that do read it. For any assistant, plugin, or agent that fetches llms.txt, you have removed the guesswork. You are telling it your bestsellers are your bestsellers, rather than hoping it infers that from your homepage layout.
3. It is future-facing at near-zero cost. The direction of travel is clear — more agents, more AI shopping, more machine-reading of stores. A file that takes an afternoon to write and a few hours to wire up is cheap insurance against that direction.

What llms.txt does not do: it does not fix a store with thin content, no schema, and vague copy. If a model follows a link in your llms.txt and lands on a product page with two lines of description and no structured data, the file has done its job and your page has failed. The citations come from the pages, not the index. That is why I always pair llms.txt work with the fundamentals — see get your Shopify store cited by ChatGPT and product schema for AI shopping.

How is llms.txt different from robots.txt and sitemaps?

This trips people up, so let me put it in a table.

The key distinction is curated versus exhaustive. Your sitemap lists every product, tag, and filtered URL — thousands of entries, most of them noise. llms.txt is the opposite: the twenty or thirty pages you would actually want an AI to read and quote. You keep all three files. They do different jobs, and llms.txt does not replace or interfere with the other two.

How do I actually add llms.txt to a Shopify store?

This is where Shopify makes you work a little, because the platform does not let you drop a raw file at the domain root the way a static host would. You cannot just upload llms.txt and have it appear at yourstore.com/llms.txt. Shopify owns the root and serves Liquid templates there.

You have a few realistic routes, from simplest to most robust:

• Redirect to a hosted file. Host the plain-text file somewhere you control (a GitHub raw URL, an S3 bucket, a Cloudflare R2 object, or a tiny static host) served as text/plain, then create a URL redirect in Shopify from /llms.txt to that URL. Fast, but it is a redirect, not a native file, so a strict fetcher that does not follow redirects will miss it.
• Serve it from an edge worker or reverse proxy. If your domain sits behind Cloudflare (or similar), a small worker can intercept /llms.txt and return the file directly with a 200 and text/plain, no redirect. This is the cleanest result: a real file at the real path.
• App proxy. A Shopify app proxy lets you serve content from your own server under your store's domain. It is the most "native" method and the right call if you also want to generate the file dynamically from live collection and product data.
• Subdomain fallback. If you cannot touch the root at all, publish at llms.yourstore.com/llms.txt or link the file from your homepage and sitemap. Less ideal — the convention expects the root — but better than nothing.

Whichever route you choose, three technical rules matter: serve it as Content-Type: text/plain; charset=utf-8, return HTTP 200 (not a soft redirect that some fetchers drop), and keep it at a predictable, stable URL. This is a couple of hours of work for a developer who knows Shopify's proxy and redirect behaviour — the kind of surgical task that sits at the bottom of my project range. There is a fuller cost breakdown in what it costs to build a Shopify website.

For reference, developer/build pricing on a job like this is small:

Those figures sit inside my standard ranges — a single feature is ₹40,000 upward, and a retainer runs ₹40,000–₹80,000 a month. The point is that the file itself is cheap; the value is in the AEO work around it.

What should go in a Shopify store's llms.txt?

Structure beats length. Keep it short and grouped. Here is what earns a place, and what does not.

Include:

• A one-line summary of what you sell and who for, right at the top.
• Your best two to five hero products — the ones you would want recommended.
• Your primary collections, grouped by how a customer thinks (concern, category, use case).
• Shipping, returns, and warranty policy pages — AI assistants get asked about these constantly.
• Sizing, care, or how-to guides that reduce buyer hesitation.
• Your about page and your contact/FAQ pages.
• One-line, plain-language descriptions on every link. No marketing fluff — describe what the page is.

Leave out:

• Filtered and sorted URLs (?sort_by=, tag combinations). They are noise.
• Thin collection or tag pages with a handful of products.
• Anything you would be uncomfortable seeing quoted back to a customer verbatim.
• Duplicate paths to the same product.

A good llms.txt for a mid-size D2C store is rarely more than 25–40 links. If it is longer, you are dumping your sitemap in, which defeats the purpose. Write it the way you would brief a new salesperson on their first day: "here are the things people ask about, and here is where the answer lives."

If your catalogue is large and changes often, generate the file from live data — pull your top collections and bestsellers programmatically so it never goes stale. That is where an app proxy earns its keep. The descriptions can lean on your product metafields, which is one more reason to get your metafield strategy right for AI readability.

What about llms-full.txt?

There is a second, optional file: llms-full.txt. Where llms.txt is the index, llms-full.txt inlines the actual Markdown content of your key pages, so a model can read the full text without following any links.

For most Shopify stores, I would not start here. The short index gets you 80% of the benefit and is far easier to keep current. Add llms-full.txt only once your index is stable and you have a way to generate it from live content — otherwise you are hand-maintaining a large document that drifts out of date the moment you edit a product. Start small, ship the short version, and expand only if the data tells you to.

Where does llms.txt fit in a wider AEO strategy?

This is the part that matters most, and the part most "just add llms.txt" advice skips. The file is one brick, not the wall. Getting a D2C store surfaced and cited by AI assistants — answer engine optimisation (AEO) and generative engine optimisation (GEO) — is a stack, and llms.txt sits near the top of it:

1. Clean, answer-shaped content. Product and policy pages that answer real questions in plain language. Without this, nothing else works.
2. Structured data (schema). Product, FAQ, and organisation schema so machines parse your pages unambiguously. This is the single highest-leverage layer for ecommerce.
3. Metafields and machine-readable attributes. Consistent, structured product data that models can extract.
4. llms.txt. The curated index that points AI at the above, once the above is worth pointing at.
5. Off-site presence. Reviews, mentions, and citations elsewhere that models cross-reference.

Notice llms.txt is layer four, not layer one. If you publish the file before doing the on-page work, you are handing a model a beautifully organised menu that leads to empty plates. Do the content and schema first; the index amplifies good pages, it does not create them.

If you want the full map of how these terms relate, I break it down in AEO vs SEO vs GEO, and the practical playbook for showing up specifically in the assistants people use is in how D2C brands show up in ChatGPT and Perplexity.

Do this in the right order

If you take one thing from this: llms.txt is genuinely worth publishing, but it is the last 10% of an AEO project, not the first. Write clean pages, add schema, structure your metafields, then publish a tight, curated index that points AI assistants straight at your best work. Skip the foundation and the file is decoration.

For a store that already has the fundamentals in place, adding llms.txt is an afternoon of work with real upside as AI shopping grows. For a store that does not, the file is a signpost to problems — which is useful information, just not the win you were hoping for.

What's next

If you want AI assistants to actually understand and quote your store, the fastest path is a proper AEO pass: clean the pages, add schema, structure the data, then wire up llms.txt so it points at work worth pointing at. Tell me your store URL, your catalogue size, and where your customers are, on the Shopify developer page or through the short intake at /start, and I will give you an honest fixed price and tell you where your gaps actually are — sometimes the file is the least of it.

Related reads

• Get your Shopify store cited by ChatGPT — the on-page work that earns citations.
• Product schema for AI shopping — the highest-leverage layer for ecommerce AEO.
• AEO vs SEO vs GEO — how the acronyms fit together, without the hype.

Sources

• llms.txt proposal — llmstxt.org
• Shopify Help Center — URL redirects
• Google Search Central — structured data guidelines