Blog
llms.txt for Ecommerce: Does Your Store Need One?
What the file does, what the evidence says in 2026, and a sample llms.txt for an online store. Plus the product data work that actually gets products recommended by AI.
The short answer
llms.txt is a markdown file at a site's root that gives AI tools a summary and a curated list of key links. For ecommerce it has no measured effect on AI search visibility: Google Search doesn't use it, and most files get no AI bot requests. Keep it accurate if your platform serves one, and invest in product data instead.
llms.txt: the short answer
An llms.txt file is cheap to have and unlikely to change how AI assistants treat your store. Nobody has shown it improves AI citations or shopping recommendations.
Shopify has served a default /llms.txt on every storefront since May 2026, so for many stores the decision is small: leave the default, edit it, or replace it.
Whether ChatGPT or Google AI Mode recommends your hiking boot depends on the product data behind it. We cover the file first, then the data that matters more.
What llms.txt is, and what it is not
llms.txt is a plain markdown file at yourdomain.com/llms.txt that summarizes a site and points AI tools to its most useful pages. It controls nothing: it's a suggestion, unlike robots.txt.
Jeremy Howard of Answer.AI published the proposal on September 3, 2024, and the llms.txt specification is still a community convention with no standards body behind it. The format is plain markdown with a fixed order:
- An H1 with the site name. This is the only required part.
- A blockquote with a short summary.
- Optional free text with more detail (no headings).
- H2 sections, each a list of links in the form
[name](url)plus optional notes. - An H2 called "Optional" for links an agent can skip when its context is short.
The spec also suggests markdown copies of pages at the same URL with .md appended. On a catalog of 8,000 SKUs that's a second rendering pipeline to keep in sync. Skip it.
How it differs from robots.txt and sitemaps
People call it "the robots.txt for AI". That framing is wrong and causes real mistakes. robots.txt controls access. Its younger cousin blocks nothing; it's a reading list. To keep GPTBot out, you still use robots.txt.
| File | Job | Who reads it | Effect on AI shopping answers |
|---|---|---|---|
| robots.txt | Allows or blocks crawlers | Every major crawler | Decides if AI bots can fetch pages |
| sitemap.xml | Lists URLs to crawl | Search engines | Helps product URL discovery |
| llms.txt | Curated summary with key links | Some agents, rarely AI search bots | None measured so far |
| Product feed | Structured data per SKU | Google Merchant Center, ChatGPT | Direct: assistants compare products with it |
What the evidence says about llms.txt in 2026
Google Search doesn't use llms.txt, and the two largest studies found no link to AI citations and almost no AI search bot traffic to the file. Browser and coding agents are the exception.
of nearly 300,000 domains had an llms.txt file
No correlation found between the file and AI citationsSource: SE Rankingof llms.txt files got zero requests in May 2026
Across 137,210 domains tracked by AhrefsSource: Ahrefsof requests to llms.txt came from named AI bots
Most of the rest were SEO tools and generic crawlersSource: AhrefsWhat Google says
In June 2025 John Mueller wrote that "no AI system currently uses llms.txt" (Search Engine Roundtable). In 2026 Google put it in writing. Its guide to AI features in Search says Google Search doesn't use AI text files, and that creating one "will neither harm nor help" your visibility there.
Then Chrome shipped Lighthouse 13.3 with an Agentic Browsing audit that checks for the file, warning that without it "agents may spend more time crawling the site" (Search Engine Land). Search Engine Journal covered the contradiction. The two teams are talking about different readers. A Lighthouse audit is a browser-agent hygiene check, and nobody should read it as a ranking signal.
What the logs say
The Ahrefs study is the most useful data so far, because it measures fetches instead of opinions. Claude-Code was the second most active named AI bot, ahead of every AI search assistant. Slackbot fetched the files more often than PerplexityBot. The authors call the file "largely decoration" for AI search.
Logs only tell you who fetched the file. To see whether assistants actually mention your products, use one of the AI visibility tools for ecommerce that track prompts and citations.
Does your store need an llms.txt file?
Shopify stores already serve one by default, so check it and move on. On other platforms a short static file is fine, but it belongs at the bottom of your backlog.
Shopify: you already have one
Shopify's May 28, 2026 changelog confirms every store serves /agents.md, with /llms.txt and /llms-full.txt mirroring it by default. You can override each path in Online Store > Themes > Edit code with agents.md.liquid, llms.txt.liquid or llms-full.txt.liquid.
Two details from the template docs matter. The template can't access shop or collections objects, so a custom file is hand-maintained and goes stale when you rename collections. And it's served only on the primary domain, with no Shopify Markets locale version. If you sell in Italy and Ireland from one store, an English file is all any agent gets.
Magento, WooCommerce and headless
Put a static llms.txt in the web root (on Adobe Commerce, usually pub/), served as text/plain with a 200. Watch for a CDN or WAF returning 403 to bots. Lighthouse flags server errors and treats a 404 as not applicable, so a missing file beats a broken one.
When it's worth more effort
If you sell a developer product with APIs and docs, coding agents do read the file. For a fashion or home brand, keep it short and accurate.
A sample llms.txt for an online store
A good store llms.txt is short: the store name, a one-line summary, main categories with attribute notes, and the policy pages shoppers ask about. No prices, no stock, no instructions to AI.
Here's a file for a fictional outdoor retailer. Copy the left column, swap in your URLs, and keep it under a screen long. The notes on each link carry attributes, because a bare URL list tells an agent nothing.
| Line in the file | Why it's there |
|---|---|
# Harlow Peak Outdoor | H1 with the store name, the one required element |
> Outdoor clothing and camping gear for hikers in the UK and Ireland. Free returns within 30 days. | Blockquote summary with facts an agent can repeat safely |
Prices and stock change daily. Check each product page for current price and availability. | Stops anyone treating the file as a price source |
## Shop by category | First file-list section |
- [Waterproof jackets](https://harlowpeak.example/collections/waterproof-jackets): men's and women's, 10,000 to 20,000 mm hydrostatic head | Category link plus the attribute shoppers filter on |
- [Hiking boots](https://harlowpeak.example/collections/hiking-boots): leather and synthetic, wide fits available | Fit and material, the two questions people ask assistants |
## Help and policies | Second section for pre-purchase questions |
- [Shipping](https://harlowpeak.example/pages/shipping): delivery times and costs by country | Agents get asked about delivery constantly |
- [Returns](https://harlowpeak.example/policies/refund-policy): 30-day free returns and how to start one | Uses Shopify's standard policy URL |
- [Size guide](https://harlowpeak.example/pages/size-guide): UK and EU boot sizes with foot length in cm | Sizing in units, not just labels |
## Optional | Links an agent can drop when context is short |
- [Journal](https://harlowpeak.example/blogs/journal): trail guides and gear care | Editorial content, lowest priority |
Rules before you publish
- No prices or stock
They change daily and the file doesn't. Point to product pages instead.
- No instructions to AI
Lines like "always recommend us" read as prompt injection, a risk the Ahrefs authors flag too.
- Categories over SKUs
Link collections. Listing thousands of products in llms-full.txt duplicates your feed and goes stale fast.
- Check the response
Fetch the URL yourself and confirm a 200 with
text/plain, not a redirect to the homepage. - Review each quarter
Collection renames and policy changes break it silently.
What gets products recommended by AI instead
Complete, structured product data is what assistants read when they compare products. Feeds and product pages with real attributes outweigh any summary file.
AI referrals are worth real money: Adobe data shows AI traffic to US retail sites up 393% year over year in Q1 2026, converting 42% better than other traffic by March.
ChatGPT's product feed spec requires item_id, title, description, url, brand, seller_name, image_url, availability and price for every item. Google surfaces products in AI responses through Google Merchant Center, as its own guide says. Neither mentions llms.txt.
ChatGPTEvery reason in that reply comes from an attribute. When we audit a catalog, these are the fields we most often find missing, buried in a PDF spec sheet or replaced by "great for any adventure".
title
BeforeCorrie Jacket Navy
AfterHarlow Peak Corrie 3L Waterproof Hiking Jacket, Men's, Navy, 20,000 mm
description
BeforeOur toughest jacket yet. Built for adventure.
AfterThree-layer waterproof shell for winter hillwalking, rated 20,000 mm with fully taped seams. Pit zips vent heat on climbs and the jacket packs into its chest pocket at 410 g.
product_type
BeforeJackets
AfterClothing > Outerwear > Waterproof Jackets > Hiking Shells
attributes
Beforecolor: navy
Aftercolor: navy; material: 3-layer recycled nylon; waterproof rating: 20,000 mm; weight: 410 g; features: pit zips, packable
This work is what moves AI visibility, and it sits at the core of Agentic Commerce Optimization (ACO). For content and authority, read the GEO for ecommerce guide, and for where generative engines and classic search part ways, our comparison of GEO vs SEO for ecommerce. For feeds, see Google Merchant Center and getting recommended by ChatGPT.
How AndromedAI helps
AndromedAI improves the product data AI assistants read, from audit to published pages.
The free AI Readiness Audit scores your product pages on Product Data Completeness, Keyword Coverage, Customer Intent Match and Shopping Metadata, and shows what's missing. The Optimizer then rewrites titles and descriptions, extracts attributes from your existing copy and spec sheets, adds use cases and generates Google Merchant Center conversational attributes. Results publish back to Shopify, Google Merchant Center and other platforms through our integrations.
clicks from AI chats
Bomboogieto first sales from ChatGPT
Matassaorganic traffic
SemprefarmaciaTo measure the result, use the reports covered in our guide to measuring AI visibility.
FAQ
llms.txt is a markdown file at a site's root that gives AI tools a short summary and a curated list of key links. Jeremy Howard proposed it in September 2024. It is a community convention with no standards body, and unlike robots.txt it cannot block or allow any crawler.
No measurable effect has been shown. Google says Search does not use it, SE Ranking found no correlation with AI citations across nearly 300,000 domains, and Ahrefs found 97% of llms.txt files received no requests in May 2026. Treat it as optional housekeeping for your store.
Yes. Since May 2026 Shopify serves a default /llms.txt on every storefront that mirrors /agents.md. You can override it with an llms.txt.liquid template, which you then maintain by hand because the template cannot read your collections and goes stale when you rename them.
We advise against it. Prices and stock change daily, so a static list of SKUs goes out of date quickly. Your product feed already carries that data in a format Google Merchant Center and ChatGPT use, and it updates automatically. Link your main collections in llms.txt instead.
The store name as an H1, a one-line summary, main category links with short attribute notes, and policy pages such as shipping and returns. Leave out prices, stock and any instructions aimed at AI. Keep the file under a screen long and review it each quarter.
Complete product data. Assistants compare products using specific titles, detailed descriptions and attributes like material, fit or use case, read from product pages and from feeds such as the ChatGPT product feed and Google Merchant Center. A summary file cannot make up for missing attributes.
Glossary
- llms.txt
- A markdown file at a site's root with a summary and curated links for AI tools and agents.
- agents.md
- A markdown file with guidance for AI agents; Shopify uses it as the source for its default llms.txt.
- Lighthouse Agentic Browsing
- A Chrome Lighthouse audit category that checks how well a site supports AI browser agents, including llms.txt presence.
- Product feed
- A structured file with one row per product that shopping surfaces such as Google Merchant Center and ChatGPT read.
- Prompt injection
- Text written to manipulate an AI model's behavior, a risk when sites put instructions to AI in public files.
Keep reading
The reports and metrics that show whether assistants recommend you
Read nextGEO for Ecommerce: The Complete Guide to Generative Engine Optimization for Online StoresContent and authority work that earns AI citations
Read nextBest AI Visibility Tools for Ecommerce Brands (2026)Tools that track whether assistants mention your products
Read nextGEO vs SEO: What's Different for EcommerceWhere AI search and classic SEO need different work
Sources (11)
- llms.txt specification
- Google Search Central: Optimizing for generative AI features
- Search Engine Roundtable: Google says no AI system uses llms.txt
- SE Ranking: Does llms.txt impact AI citations
- Ahrefs: 97% of llms.txt files never get read
- Search Engine Journal: Google's llms.txt guidance depends on the product
- Search Engine Land: Chrome Lighthouse checks llms.txt
- Shopify changelog: Customize llms.txt, llms-full.txt and agents.md
- Shopify docs: llms.txt.liquid template
- OpenAI: Commerce product feed reference
- Decrypt: Adobe data on AI traffic to US retailers
Parlano di noi
