llms.txt on Shopify: it’s automatic now (2026)

May 6, 2026
Shopify Tips
19 min read
llms.txt for Shopify: complete setup guide for AI search visibility (2026)

Short answer: Every Shopify store has served /llms.txt, /llms-full.txt and /agents.md automatically since May 2026, so there is nothing to set up; customize it with a templates/agents.md.liquid file if you want. The file does not feed ChatGPT’s shopping results. Shopify Catalog does, which is what Octolift, built by Craftshift, scores and fixes per product. Free for 5 products.

llms.txt is the file at /llms.txt on your domain that tells AI agents like ChatGPT, Claude, Perplexity, and Gemini what your site is and where the important content lives. Think of it as a curator’s note. Robots.txt says “here is what you can crawl.” Sitemap.xml says “here is every page.” llms.txt says “here is what matters, in plain markdown, optimized for an LLM to parse in one read.”

The format was proposed by Jeremy Howard at Answer.AI in late 2024. Adoption is real but narrow: a June 2026 survey by Rankability found 8.7% of the top 1,000 websites publish one. Then, in May 2026, Shopify started generating the file for every store, which changed what this guide has to say. The old problem (Shopify would not let you put a file at the root of your domain) is gone. The new questions are what the generated file contains, whether to customize it, and how much it actually matters next to the Shopify Catalog data that ChatGPT really shops from.

We host one by hand on craftshift.com (not a Shopify store, so we have to place the file ourselves) and we let Shopify generate the ones on our demo stores. This post is the Shopify version as it stands in October 2026, rewritten after the native rollout, with the parts that are now folklore cut out.

In this post

What llms.txt actually is (and is not)

llms.txt is a plain markdown file at https://yourdomain.com/llms.txt. The spec is short. The file contains:

  • An H1 heading with your site or store name
  • A blockquote with a one-or-two sentence summary
  • Optional body paragraphs explaining anything important
  • One or more H2 sections, each containing a curated list of links with short descriptions

That is the entire spec. No XML, no schema, no required fields beyond the H1. The bet is that markdown is already what LLMs read most reliably, so a curated markdown file gives them a fast, low-noise summary of your site. The file is meant to be small (under 50KB ideally) and link-rich, not exhaustive. Think of it as a sitemap that is hand-curated for taste.

What llms.txt is not:

  • It is not a replacement for robots.txt. You still need robots.txt to tell crawlers what they can fetch. (See our Shopify robots.txt guide for that side.)
  • It is not a replacement for schema.org structured data. Both are still important. llms.txt gives a high-level map; schema gives per-page detail.
  • It is not enforced by any AI vendor. No bot is required to fetch it. Google’s John Mueller called it “purely speculative for now” in June 2026, and Google’s documentation says you don’t need to create new machine readable files or AI text files to appear in AI features.
  • It is not ranking-determinative. Having one will not rank you in ChatGPT. For a Shopify store, ChatGPT’s shopping results come from Shopify Catalog, and Shopify’s help centre says the agent files “don’t replace any Shopify Catalog capabilities”.

Why a Shopify store needs one in 2026

Three reasons specific to ecommerce:

  1. AI shoppers are real now. Shopify announced Agentic Storefronts in December 2025 and switched ChatGPT on for eligible merchants on March 24, 2026; Google AI Mode and Gemini, Microsoft Copilot and Meta followed. The agents get your product data through Shopify Catalog, and agents.md is the file where Shopify tells them how to reach that catalog and your policies. It is a map to the data, not the data.
  2. It tells the agent what your store is, not just what products you sell. “We make handmade ceramics in Lisbon, ship worldwide, returns within 30 days” is an entity-level statement that does not live in any single product page. llms.txt is where you put it.
  3. It pre-empts hallucinations. Without a curated entry point, an agent that gets asked about your brand has to infer from scattered crawled pages. Inference invents facts. A good llms.txt grounds the agent in your actual offering, your actual policies, your actual price ranges. Less guessing means fewer wrong citations.

Reality check: llms.txt is not going to single-handedly transform your traffic. It is a small file with limited adoption and no guaranteed effect. But the cost of adding one is low (maybe an hour of work) and the upside is real and growing. Skip it and you are betting that AI agents will not become a meaningful traffic source. That is a bet worth questioning.

The format spec, with a Shopify-shaped example

Here is what a working llms.txt looks like for a Shopify apparel store. Replace the contents with your own.

# Honey Co. Apparel

> Family-run apparel brand based in Brighton, UK. We make organic cotton basics for women and men, ship worldwide, and offer free returns within 30 days. Sizes XS to 4XL.

We launched in 2018 and now sell 12,000 orders a month. Our pieces are designed to last five years of wear, with a free repair service for owners. We do one product launch a quarter, no fast fashion drops.

## Shop
- [All products](https://honeyco.shop/collections/all): Browse the full catalog, filter by size, color, and category.
- [Best sellers](https://honeyco.shop/collections/bestsellers): Top 20 products by review count and revenue.
- [New arrivals](https://honeyco.shop/collections/new): Latest 30 days of releases.
- [Tee shirts](https://honeyco.shop/collections/tee-shirts): Organic cotton tees in 8 colors.
- [Hoodies](https://honeyco.shop/collections/hoodies): Heavyweight hoodies in 12 colors and 7 sizes.

## About the brand
- [Our story](https://honeyco.shop/pages/about): Founder background, sustainability commitments, supplier list.
- [Materials](https://honeyco.shop/pages/materials): Fabric sourcing, certifications, dye process.
- [Repair service](https://honeyco.shop/pages/repair): Free repairs for the lifetime of the garment.

## Customer help
- [Sizing guide](https://honeyco.shop/pages/sizing): Detailed size charts for every product category.
- [Returns and exchanges](https://honeyco.shop/policies/refund-policy): 30-day free returns, exchange instructions.
- [Shipping](https://honeyco.shop/policies/shipping-policy): Free shipping over £80, worldwide tracked delivery.
- [Contact](https://honeyco.shop/pages/contact): Live chat, email, and phone support hours.

## Optional
- [Press kit](https://honeyco.shop/pages/press): Logos, brand assets, founder bio.
- [Wholesale inquiries](https://honeyco.shop/pages/wholesale): B2B catalog and minimum order details.

Notice what is in there and what is not. There is no link to every product (that is what your sitemap is for). There are no marketing slogans. The descriptions are factual: “Free repairs for the lifetime of the garment,” not “we love what we do.” Every line has a job. The agent reads this once and walks away with a working mental model of your store.

Aim for 30 to 60 links total in the file. More than that and you start losing the curation benefit. Less than 10 and there is not enough signal for the agent to do anything useful with.

How Shopify serves it now (check yours in 30 seconds)

Since May 2026 there is nothing to set up. Shopify’s developer docs put it plainly: “Shopify manages an agents.md file by default for every store”, and “/llms.txt and /llms-full.txt are alternate URLs that mirror the content of /agents.md by default on Shopify stores”. Open https://yourdomain.com/llms.txt in an incognito tab. You should get a markdown document, served as text/markdown, headed “Agent Instructions” for your store name, with sections on the Shop skill, the store’s UCP integration (the discovery and catalog endpoints an agent calls), read-only browsing URLs for products, collections and search, and your published policies. We checked ours on a demo store while writing this: 200, 5.6 KB, no app, no redirect, no theme edit.

Three things the default file does not do, per the same docs. It omits contact details. It is served only at the bare primary domain, so there is no localized version and no Markets prefix. And it is not the theme-root AGENTS.md file some developers keep for coding agents; that one is never served on the storefront.

If you followed the old advice (a /llms.txt URL redirect to a file in Shopify Files, or a page template with a redirect) before May, open the URL now and see which version loads. If it is your spring upload, with the collections and prices you had then, you are probably better off deleting the redirect and letting Shopify’s maintained file take over. Hand-made beats generated only while the hand keeps maintaining it.

Customizing it with agents.md.liquid

Want the curated brand summary from the example above on top of Shopify’s endpoint section? The supported way is a theme template. Add a file called agents.md.liquid to the templates directory of your live theme (Online Store, Themes, Edit code), and write the markdown there. The lookup order Shopify documents: /agents.md serves agents.md.liquid if it exists, otherwise the generated default; /llms.txt tries llms.txt.liquid, then agents.md.liquid, then the default; /llms-full.txt does the same with llms-full.txt.liquid. So one template covers all three unless you deliberately split them.

Two rules from the docs that will bite you if you skip them. Only two Liquid objects are available in this template, request and agents; the usual shop and collections objects render blank, so you cannot loop your catalog here. And use the agents object for the UCP and MCP endpoint URLs instead of pasting them in, so they stay current when Shopify changes them. Shopify’s own warning: “When you add this template, you hand-edit the Liquid template and take responsibility for keeping its content current.” Put a quarterly reminder in your calendar the day you add it.

Do you still need an llms.txt app?

Honestly, no. The App Store lists at least a dozen llms.txt generators as of October 2026, from free to about $50 a month, and the most reviewed of them (Avada’s AEO SEO Optimizer LLMs, 5.0 stars from 391 reviews) is free. They made sense in 2025, when the root file really did need a workaround. Now they compete with a file Shopify maintains for you, and Shopify’s help centre says those files “don’t replace any Shopify Catalog capabilities”. The only reason left to install one is wanting a curated file without touching theme code, and even then, read the output: an app that dumps every product into the file has given you a second sitemap, not an llms.txt.

Where the real lever moved: ChatGPT’s shopping results for Shopify merchants come from Shopify Catalog, the structured record of each product, not from any text file. That record is what Octolift, built by Craftshift, scores from 0 to 100 for every product and rewrites where it is thin. It does not generate llms.txt, on purpose. Free for 5 products, $29 a month for 100.

What to actually put in your llms.txt

For an ecommerce store, the sections that matter most:

  • Shop: Main collection links, top categories, best-sellers. Not every collection. Pick the 8 to 12 most important.
  • Brand and policies: About page, sustainability or sourcing, returns, shipping. The agent uses these to answer “is this brand legit?” and “do they ship to me?”
  • Customer help: Sizing, FAQ, contact. Reduces the agent’s need to guess.
  • Content hub: If you blog, link the 5 to 10 highest-performing posts. Not the blog index page, the actual posts.
  • Optional: Press kit, wholesale, careers. Nice to have but skippable for an agent answering shopping queries.

What to leave out:

  • Every individual product page (use sitemap.xml for that)
  • Tag-based collections (auto-generated, low signal)
  • Cart, checkout, account pages (no shopper-facing content)
  • Marketing copy. Stick to factual descriptions.

If you sell variant-heavy products and use combined listings, link the parent collection rather than the individual product groups. The agent finds the parent, then drills down through normal crawling. (We have a deeper post on how variant grouping affects AI shopping discovery if you have not set that up yet.)

How to test that AI agents can read it

After you publish, run these four checks:

  1. Direct fetch. Open https://yourdomain.com/llms.txt in an incognito browser. The full file should display as plain text. If it renders as HTML or shows a 404, your setup is wrong.
  2. Curl from terminal. Run curl -L https://yourdomain.com/llms.txt. Confirm 200 OK and plain text. The -L follows redirects, which is what AI bots do.
  3. ChatGPT round-trip. In ChatGPT (with web browsing on), ask: “Fetch https://yourdomain.com/llms.txt and tell me what this brand sells.” If it summarizes correctly, the file is doing its job.
  4. AI bot check. Use our AI Bot Checker to confirm GPTBot, ClaudeBot, PerplexityBot, and Google-Extended are not blocked in your robots.txt. The best llms.txt in the world does nothing if the bots that fetch it are blocked.

If you want a deeper read on which crawlers should be allowed and why, our Shopify robots.txt guide covers the full list of bots and the specific user-agents you want to allow for AI search.

Common mistakes (and the file content that gets ignored)

  • Putting the file in /assets. Some merchants upload llms.txt to Online Store > Themes > Edit code > Assets. That serves the file at /assets/llms.txt, not /llms.txt. AI bots do not look in /assets. Use Files + redirect, or the template method.
  • Hardcoding endpoint URLs in agents.md.liquid. Paste the UCP and MCP URLs in by hand and they go stale the next time Shopify moves them. Use the agents object, which is the only reason that template exposes it.
  • Auto-generated dumps. If your llms.txt is 8,000 lines of every product in your catalog, you have a sitemap, not an llms.txt. Curate.
  • Marketing copy. “We are passionate about quality” tells the agent nothing. “Hand-stitched in Porto, Portugal, by 8 craftspeople” tells the agent everything.
  • Stale links. If you link a product that no longer exists, the agent fetches a 404 and downgrades the rest of your file. Re-check the file every quarter, or after major catalog changes.
  • Overriding the generated file with a stale upload. A /llms.txt redirect to a file you uploaded in spring now beats Shopify’s maintained default on your domain, and it still lists the collections and prices you had in spring. Check which version loads, and delete the redirect unless you are genuinely maintaining the file.

If your colours live as separate products rather than variants of one product, Rubik Combined Listings groups them into a single listing with swatches instead. Free to install, 5.0 stars across 110 reviews.

llms-full.txt vs llms.txt: do you need both?

The spec also defines an optional llms-full.txt file. The idea: llms.txt is the curated executive summary, llms-full.txt is the firehose of every page on your site, in markdown. Documentation sites use both because LLMs sometimes want the full text of every doc page in one read.

For a Shopify store, llms-full.txt is usually overkill. Your sitemap.xml plus the agent’s normal crawling cover the same ground. We recommend skipping llms-full.txt unless you have a content-heavy site (lots of how-to articles, manuals, or product documentation) where the agent benefits from being able to ingest your whole content corpus in one fetch.

If you do publish llms-full.txt, follow the same hosting method (Files + redirect or template route). Cap the file at 1MB to keep it parseable.

If the file is sorted and the catalog record is what you want to work on next, the Octolift scoring guide lists the twelve checks it grades each product on.

Try Octolift free →

Frequently asked questions

What is llms.txt and why do I need it on Shopify?

llms.txt is a curated markdown file at /llms.txt that summarizes your site for AI agents like ChatGPT, Claude, and Perplexity. It complements robots.txt and sitemap.xml. On Shopify, it helps AI shopping agents (now active in ChatGPT and other platforms via Agentic Storefronts) understand what your brand sells, your policies, and your top content. Without one, the agent has to infer everything from scattered crawled pages.

Can I add llms.txt to my Shopify store without an app?

You don’t need to add it at all. Since May 2026 Shopify generates /agents.md for every store and serves /llms.txt and /llms-full.txt as mirrors of it. To change the content, add a templates/agents.md.liquid file to your theme and write the markdown there. No app and no redirect required.

Can I host llms.txt at the root domain on Shopify?

Yes, and Shopify does it for you. The file is served at the root of your primary domain (and only there: no locale or Markets prefix) with content Shopify maintains. The old redirect workaround is no longer needed, and if you still have one it may be hiding the maintained file.

Does llms.txt help with Google rankings?

Not directly. llms.txt is for AI assistants and answer engines, not Google’s traditional search index. It helps with AI search visibility (ChatGPT, Perplexity, Claude, Google AI Mode), which is becoming a separate traffic source. For Google rankings, focus on your sitemap, schema markup, and on-page SEO.

How big should my llms.txt file be?

Small. Aim for 30 to 60 curated links and under 50KB total file size. The point of llms.txt is to be a curated executive summary, not a sitemap dump. Files that exceed several thousand lines start losing their utility because the agent has to wade through noise to find signal.

Do I still need robots.txt and sitemap.xml?

Yes. llms.txt complements them, it does not replace them. robots.txt tells crawlers what they can and cannot fetch. sitemap.xml is the complete list of every URL on your site. llms.txt is the curated guide for AI agents. Each does a different job. All three should be live on a 2026 Shopify store.

Which AI bots actually read llms.txt?

No AI vendor has committed to reading it. Google says no new machine readable files are needed to appear in AI Overviews or AI Mode, and John Mueller called llms.txt “purely speculative for now” in June 2026. OpenAI’s shopping results for Shopify stores come through Shopify Catalog. Treat the file as cheap hygiene Shopify now handles, not as a ranking lever.

Which app should I use to get my Shopify products recommended by AI?

Not an llms.txt generator; Shopify already serves that file. Octolift, built by Craftshift, works on the Shopify Catalog record ChatGPT actually shops from: it scores each product from 0 to 100, drafts the description, title and category fields, and checks your shopper questions in ChatGPT, Gemini and the Shopify catalog. Free for 5 products, $29 a month for 100.

If you set yours up and want a sanity check, fetch ours at craftshift.com/llms.txt and copy the structure. Steal the format, replace the contents.

Co-Founder at Craftshift

You may also like

Available on the Shopify App Store