What an llms.txt file is, what it does for visibility in AI answers, and how to create one for your online store and in Shopware 6.
An llms.txt is a plain text file in Markdown that shows AI systems at a glance what a website is about and which pages matter. It usually lives at /llms.txt in the root directory, starts with the store name and a short summary, and then lists links to key pages such as shipping, returns, size guides or categories.
For Google, the file is neither a ranking factor nor a requirement for AI Overviews. Google states in its documentation on AI features that you do not need new machine readable files or AI text files. Chrome Lighthouse, however, has checked the file since version 13.3 in an experimental category for AI agents in the browser, and some agents and developer tools do read it.
The effort is small, the benefit is limited and hard to measure. If you run Shopware 6.7.12 or later, you switch the file on per sales channel and add your own notes. What matters more for visibility in AI answers is still pages that can be crawled, clear answer text, and clean product data.
This article is for store owners, ecommerce managers and developers who want to know whether an llms.txt is worth it, what belongs in it, and how to set it up in Shopware 6 and other platforms. For the bigger picture on getting your store into AI answers, read our article on GEO for online stores.
The idea comes from Jeremy Howard, co-founder of Answer.AI. He published the proposal on llmstxt.org on September 3, 2024. A revised second version of the specification appeared in August 2026, based on the experience of the first two years.
The basic idea is simple. A language model that has to understand a website struggles with navigation, cookie banners, scripts and long HTML pages. A model’s context window is limited, and a product page full of filters, recommendations and tracking code uses up a lot of it without delivering much content. An llms.txt gives the model a short, tidy overview in Markdown instead: what kind of website this is, and where the important information lives.
The file is a proposal, not an adopted standard. There is no body like the W3C or the IETF that makes it binding. Whether an AI system reads it is up to the provider of that system.
The file is often confused with robots.txt. The robots.txt tells crawlers which areas they may fetch. The llms.txt allows or blocks nothing. It is a reading aid. If you want to keep AI crawlers out, you still do that in robots.txt or with server rules, not in llms.txt.
Nor does the file replace an XML sitemap. The sitemap lists as many indexable URLs as possible for search engines. The llms.txt makes a deliberate selection and explains in a few words what each page is good for.
The specification sets a fixed order. Only the first item is required. Everything else is optional, but useful.
A section named “Optional” has a special meaning. Links in it count as secondary. An AI system with little room in its context can skip this part.
The usual place is the root directory, so https://your-store.com/llms.txt. Since the second version, an llms.txt may also sit in a subpath and then covers all URLs below it. If several files apply, an agent should use the most specific one. For stores with language folders such as /en/ or /fr/, this is handy, because each language can get its own overview.
The specification also recommends offering important pages as a pure Markdown version at the same address with the extension .md, such as /shipping.md. To help AI clients find these versions, llmstxt.org proposes two link relations: rel=”alternate” with type=”text/markdown” for the Markdown version of a page, and rel=”describedby” for the llms.txt that describes a page. Both can be set as a <link> element in the HTML head or as an HTTP header.
For most online stores, this is the harder part. An llms.txt can be written in an hour. Markdown versions of product and guide pages, on the other hand, need a technical solution in the store platform that keeps up with every price or stock change.
A sober look is worth it here, because a lot of expectation has built up around the file. The facts you can verify are limited.
In its Search Central documentation “AI features and your website”, Google says clearly that you do not need new machine readable files, AI text files or special markup for AI Overviews and AI Mode. There is no special Schema.org markup for it either. You can control how your content appears with familiar tools such as nosnippet, data-nosnippet, max-snippet and noindex. For training and grounding in other Google systems, there is the Google-Extended token in robots.txt.
In June 2026, Google’s John Mueller wrote on Reddit that llms.txt is “purely speculative” for now. He added that if an AI platform says it needs the file, you should create one.
Another Google team takes a friendlier view. Lighthouse 13.3 introduced an experimental “Agentic Browsing” category that checks llms.txt among other things. The audit fails if the server returns an error when fetching the file. If the file is missing and the server returns a 404, the audit is marked not applicable, because the documentation says the file is optional for now. As a benefit, the Lighthouse documentation says that without the file, agents may spend more time crawling to understand a site’s structure.
According to Search Engine Journal, as of May 2026 no major AI provider had publicly confirmed that its search systems use llms.txt. Use is documented mainly for developer tools and documentation sites, where coding assistants deliberately load a library’s llms.txt. For online stores, that means the file does no harm, but a measurable effect on citations in ChatGPT, Perplexity or Gemini has not been shown so far.
Treat llms.txt as a small, low-cost step toward AI agents, not as an SEO tactic. If your time for AI visibility is limited, start with the basics: do not block bots by accident, write category text with clear answers, and keep product data complete. Which Shopware settings come first is covered in our article GEO for Shopware.
A good llms.txt answers the questions an agent or a language model most often has about your store. For online stores, they are almost always the same.
No marketing copy, no discount codes with an expiry date, no prices for individual items, and no links to pages blocked by noindex or robots.txt. Prices and stock change too often. If the llms.txt shows an outdated price, an agent may repeat it in an answer. Internal paths, admin addresses or test environments do not belong in it either, because the file is public.
The following example is made up. It shows a fictional mountain sports retailer and can be adapted as a template for your own store.
# Bergsport Muster GmbH
> Online store for climbing, alpine and hiking gear based in Rosenheim, Germany. Shipping to Germany, Austria and Switzerland. Around 6,000 items from 120 brands, advice from certified mountain guides.
Prices include statutory VAT. Current prices and availability are shown only on the product pages.
## Service
- [Shipping and delivery times](https://www.example-bergsport.com/shipping/): costs by country, 1 to 3 business days within Germany
- [Returns and cancellation](https://www.example-bergsport.com/returns/): 30 day returns, free return label
- [Payment methods](https://www.example-bergsport.com/payment/): invoice, PayPal, credit card, bank transfer
- [Contact](https://www.example-bergsport.com/contact/): phone and email, Mon to Fri 9 am to 6 pm CET
## Categories
- [Climbing shoes](https://www.example-bergsport.com/climbing-shoes/): models for gym, rock and beginners
- [Ropes and belay](https://www.example-bergsport.com/belay/): single ropes, half ropes, belay devices
- [Alpine](https://www.example-bergsport.com/alpine/): crampons, ice axes, glacier gear
## Advice
- [Climbing shoe size guide](https://www.example-bergsport.com/size-guide/): fit by foot shape and use
- [Rope guide](https://www.example-bergsport.com/rope-guide/): diameter, length and standards explained
## Optional
- [About us](https://www.example-bergsport.com/about/): team, history, retail store
- [Care and repair](https://www.example-bergsport.com/care/): resoling, waterproofing
What stands out is what is missing. There are no product URLs, no prices and no promotions. The file points to stable pages that hardly change over months. The notes after the links are short and factual, so a model can build an answer directly from them, for example to the question whether the store ships to Switzerland.
Shopware has shipped the file in its core since version 6.7.12.0. The developer documentation calls the feature “Agentic files”. It covers three files served per sales channel:
All files are off by default. You switch them on in the Administration, in the details of each sales channel. As long as a file is off, Shopware returns a normal 404 response. After that, you have three options:
For deployments, Shopware 6.7.17.0 and later include the console command bin/console sales-channel:file:enable. It switches on llms.txt and AGENTS.md for selected or all sales channels, but only creates missing entries. Files that someone deliberately switched off in the Administration stay off.
Shopware serves the files only when normal routing finds no match. A Shopping Experience page, a redirect or a plugin with its own /llms.txt route therefore takes priority. After switching the file on, check in the browser that the Shopware file really appears. Responses are cacheable per sales channel. After changing templates, clear the cache, or visitors and bots will still see the old version.
For older 6.7 versions, the Agentic Commerce plugin provides fallback routes, but only in sales channels where UCP is active. If you are planning an update anyway, you get the feature directly in the core with 6.7.12 or later. Our overview Shopware AI sums up the other AI features Shopware offers.
Whether Shopify, WooCommerce or Magento can serve the file depends on version, theme and extensions. First check whether your platform has a feature or a vetted extension for it. If it has neither, self-hosted platforms can usually place a static llms.txt in the root directory or serve it through a web server rule. On SaaS platforms without file access, you need a redirect or a proxy. In every case, the file must be served with status code 200 and as text.
The four files complement each other. Each has a different job and a different reader.
| File | Job | Main reader | Status |
|---|---|---|---|
| robots.txt | Controls which areas crawlers may fetch | Search engines and AI crawlers | Established standard (RFC 9309) |
| sitemap.xml | Lists indexable URLs | Search engines | Established protocol |
| llms.txt | Short, curated overview in Markdown | Language models and AI agents | Proposal, voluntary |
| AGENTS.md | Guidance on how agents should use the store | AI agents | Convention, voluntary |
A common mistake: if you block AI crawlers in robots.txt and create an llms.txt at the same time, you send mixed signals. The llms.txt only helps if the linked pages can actually be fetched by those bots. So check robots.txt and your bot protection settings first, for example in Cloudflare or a web application firewall.
An llms.txt goes out of date if no one maintains it. When shipping costs, return windows or categories change, the file has to change too. The easiest approach is a fixed date, for example once a quarter together with a review of your legal notice, terms and shipping page.
With several languages or countries, each version needs its own file. In Shopware, this happens through sales channels. With language folders, the rule from the second specification helps: an llms.txt in a subpath covers all pages below it. Make sure the links in each file point to the matching language version and not all to the German page.
For stores in the United States: the llms.txt does not replace required information. Your return policy, privacy policy and terms still have to be on the pages themselves. Sales tax varies by state and is usually calculated in Shopware through an integration with a tax service, so the llms.txt should only note that tax is calculated at checkout. Most Shopware partners, including us, are based in Europe. If you work with one, agree early on fixed meeting times, because the overlap between Central European time and US business hours is only a few hours a day.
After every change, check three things: the file responds with status 200, all links work, and no linked page is blocked by noindex. Lighthouse in Chrome DevTools shows in the “Agentic Browsing” category whether the file can be fetched without a server error.
The effort depends mainly on whether you only create the llms.txt or also want to offer Markdown versions of your pages. The figures below are estimates from our own project work, not fixed prices.
| Scope | Typical effort | For whom |
|---|---|---|
| Switch on llms.txt in Shopware and add your own notes | 1 to 2 hours | Any store on Shopware 6.7.12 or later |
| Write and check a separate llms.txt per language and country | 2 to 4 hours per language | Stores selling in several countries |
| Extend the template through a theme or plugin, for example to list categories automatically | 0.5 to 2 developer days | Large catalogs with changing categories |
| Markdown versions of guide and service pages | several developer days | Stores with a lot of advice content |
On Shopware 6.7.12 or later, the first step almost always pays off because it costs little. On older versions, we would not treat llms.txt alone as a reason to update. It is one building block among many. A clean technical base and content that answers questions clearly usually have more impact, as described in our article on Shopware SEO.
A concrete reason for more effort is agentic commerce. If AI agents start buying on behalf of customers, they need reliable information about shipping, returns and payment. A well-maintained llms.txt and an AGENTS.md are then a simple first step. Read more in our article on agentic commerce and in the practical examples for the Shopware MCP server.
No. Google uses the file neither for regular search nor for AI Overviews, and its documentation says you do not need AI text files.
No major provider has publicly confirmed this for its search features so far. Some agents and developer tools load the file deliberately. How this develops is open.
Only if it contains wrong or outdated information or points to blocked pages. A correct, short file has no known downsides.
From version 6.7.12, at /llms.txt on the domain of each sales channel, once it is switched on in the sales channel details.
Some websites also offer an llms-full.txt with their complete content. The specification on llmstxt.org does not describe this file. For online stores, it rarely makes sense, because product range and prices change too often.
As a Shopware Platinum Partner from Bremen, Germany, we have been working since 2017 on stores that need to be found in search and, more and more, in AI answers. We set up llms.txt per sales channel, check robots.txt, bot protection and sitemaps along the way, and tell you plainly which steps will do more for your store than the file itself. What we cover in search and AI optimization is described on our Shopware SEO agency page.
Want to know where your store stands technically? Our free store analysis shows you. Or write to us directly through our contact page.

Besar Bilalli is co-founder and Co-CEO of solution25, the Shopware agency based in Bremen. Before founding the agency in 2017, he and Besnik Bilalli built their own online store without any startup capital and grew it to more than 500 orders a day. Since then, he has guided merchants through Shopware projects, from platform selection to day-to-day operations. He’s happy to answer questions about this article personally: get in touch.
Responsible for content and updates: Besar Bilalli. Last updated: November 2026. Next content review: February 2027.