What's the difference between llms.txt, a sitemap and structured data?

If you got here comparing llms.txt, sitemaps and structured data, chances are someone pitched llms.txt to you as "the sitemap for AI," or as the thing that replaces the markup on your product pages. It's neither. These are three pieces with different owners, different readers and different jobs: the sitemap tells search engines which URLs exist, structured data says what's on each page, and llms.txt suggests where agents should start reading. The llms.txt proposal frames it the same way: it's meant to sit alongside the standards that already exist, not to replace them.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2}

Who defines each one

llms.txt is a proposal, not a standard. Jeremy Howard published it on September 3, 2024, with a second version in August 2026, and its own page presents it as a proposal to standardize the use of an /llms.txt file.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2}

The sitemap is a protocol the search engines back. The format is defined by sitemaps.org, at version 0.90, published under a Creative Commons license and, according to the site itself, supported by Google, Yahoo! and Microsoft.⁠^{[4]sitemaps.org. sitemaps.org}

Structured data has two layers. The first is the vocabulary, Schema.org: a community activity founded by Google, Microsoft, Yahoo and Yandex, and run by a group with representatives from those companies, one from the W3C and a few individuals who have made significant contributions.⁠^{[5]Schema.org. About Schema.org} The second is the requirements of whoever reads it. Google, for example, treats the three possible formats equally as long as the markup is valid and follows each feature's documentation, and recommends whichever is easiest to maintain, which in most cases is JSON-LD.⁠^{[3]Google Search Central. Introduction to structured data markup in Google Search}

Who reads each one

Search engines read the sitemap, and they say so. Google describes it as a file that search engines read to crawl your site more efficiently.⁠^{[2]Google Search Central. What is a sitemap} Google, Bing and the other major search engines accept a sitemap: line in robots.txt that points to the file and isn't tied to any particular agent.⁠^{[6]Google Crawling Infrastructure. How Google interprets the robots.txt specification} Bing went further in July 2025: with AI search in Copilot under way, it called sitemaps a foundational signal for making sure every URL on your site is covered.⁠^{[10]Microsoft, Bing Webmaster Blog (July 2025). Keeping Content Discoverable with Sitemaps in AI Powered Search}

Structured data has declared readers too. Google uses the structured data it finds to understand the content of your pages.⁠^{[3]Google Search Central. Introduction to structured data markup in Google Search} Microsoft, on its advertising platform's blog, writes that schema.org markup helps search engines and AI systems understand your content.⁠^{[7]Microsoft Advertising (Krishna Madhavan). Optimizing Your Content for Inclusion in AI Search Answers}

With llms.txt, we know who it's meant for, not who reads it. According to the proposal, an AI agent consults it when it needs information, at the moment of answering; its author expected it to serve that purpose more than model training.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2} What the proposal can't tell you is who actually requests it. Google Search has already answered that it doesn't use it, as you'll see below, and the page where OpenAI documents its crawlers explains how to control them with robots.txt and never mentions llms.txt.⁠^{[11]OpenAI. Overview of OpenAI Crawlers} For your store, the answer is in your server logs.

What each one says about your store

The sitemap gives addresses. Each URL goes in its own tag, with the date it last changed if you choose to provide it. Google ignores <priority> and <changefreq>, and uses <lastmod> only if it's consistently and verifiably accurate.⁠^{[9]Google Search Central. Build and submit a sitemap} Bing doesn't use the first two either, and asks that <lastmod> reflect the last change to the page's content, not the date the file was generated.⁠^{[10]Microsoft, Bing Webmaster Blog (July 2025). Keeping Content Discoverable with Sitemaps in AI Powered Search} It says nothing about what a product page sells or at what price.

Structured data gives the content. Google defines it as a standardized format for providing information about a page and classifying its content.⁠^{[3]Google Search Central. Introduction to structured data markup in Google Search} On a product page, that means its name, its price, its currency and its availability, in named fields. And it comes with a rule the other two don't have: Google asks you not to mark up content that isn't visible to people reading the page,⁠^{[12]Google Search Central. General structured data guidelines} and, for its AI features, that structured data match the visible text.⁠^{[13]Google Search Central. AI features and your website}

llms.txt gives a route. It carries a title, a summary and lists of links to the most useful pages, ideally to clean Markdown versions of them, the idea behind Markdown for agents.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2} The proposal draws the line against the sitemap: a sitemap lists everything indexable for search engines, while llms.txt offers a selection for a model. A sitemap usually doesn't link to the versions a model reads well or to pages on other sites, and it almost always covers more than fits in a model's context window. llms.txt can also point to the structured data the site uses, to help interpret it.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2}

In one line: the sitemap says where; structured data, what; llms.txt, where to start.

What Google says about llms.txt

Google has put it in writing, in its documentation. On May 15, 2026, it published its guide to optimizing for generative AI features on Search, with a section that busts myths about AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization), and on June 15, 2026, it added a note on llms.txt files.⁠^{[14]Google Search Central. Latest documentation updates} The guide says:

"You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them."⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}

And a sentence later:

"It's completely fine if you decide to create and maintain LLMS.txt files (or other similar files) for other services or systems that use these files. Doing so will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}

Elsewhere in the same guide, "creating unnecessary AI text files (like llms.txt)" is listed among the tactics you can ignore for Google Search.⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search} Three things to keep in mind when you read it:

• Seeing your llms.txt crawled doesn't contradict any of this. The guide warns that Google may discover, crawl and index many kinds of files besides HTML, and that this doesn't mean it treats them in any special way.⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}
• It's about Search, not all of Google. That includes its generative AI features, such as AI Overviews and AI Mode.⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search} Other parts of Google treat the file differently: Lighthouse, Chrome's auditing tool, calls llms.txt an emerging convention and, if your server doesn't have one, marks its check as not applicable, because offering it is optional for now.⁠^{[15]Chrome for Developers. llms.txt (Lighthouse, agentic browsing audits)} And the Gemini API documentation publishes its own llms.txt.⁠^{[16]Google AI for Developers. Gemini API: llms.txt}
• Structured data gets different treatment in the same guide. It isn't required for generative AI search and there's no special schema.org markup to add, but Google says it's a good idea to keep using it as part of your SEO, because it helps with being eligible for rich results.⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}

Lined up, here's what Google Search says about the three: it ignores llms.txt, recommends structured data and treats the sitemap as a hint.⁠^{[9]Google Search Central. Build and submit a sitemap}

llms.txt vs sitemap: where each one stops

None of the three does what it's sometimes credited with.

• A sitemap doesn't index anything. It helps search engines discover your URLs, but it doesn't guarantee that all of them will be crawled and make it through indexing,⁠^{[2]Google Search Central. What is a sitemap} and submitting it is only a hint: Google doesn't even guarantee it will download it.⁠^{[9]Google Search Central. Build and submit a sitemap}
• Structured data guarantees nothing in the results. Google doesn't promise that your markup will appear in its results, even when the page is correctly marked up according to its own test.⁠^{[12]Google Search Central. General structured data guidelines}
• llms.txt neither grants nor removes permissions. Who may crawl your store is decided in robots.txt, which tells automated tools what access is considered acceptable; llms.txt is used on demand, when an agent needs information.⁠^{[1]llmstxt.org (Jeremy Howard). The /llms.txt file, v2}
• None of them tells you whether your published product page carries what you saved. That's publication verification, and how it differs from validating the markup is covered in the validate vs verify comparison.

On a Shopify store

If your store runs on Shopify, two of the three pieces already exist without you lifting a finger. Shopify generates a sitemap.xml at the root of every domain, which links to separate sitemaps for products, collections, blogs and pages, with each product's main image.⁠^{[17]Shopify Help Center. Finding and submitting your sitemap} And by default it manages a /llms.txt that mirrors the content of /agents.md, the document the platform treats as the main one for agents to discover the store; its documentation recommends against creating your own template unless you need llms.txt to say something different, because keeping it up to date then becomes your job.⁠^{[18]Shopify Dev Docs. llms.txt.liquid}

Structured data doesn't sit at a URL on the root: it lives on each product page. The guide to auditing a product page's structured data explains how to read it and compare it with what the page shows.

Pick where to start

You don't choose between llms.txt, a sitemap and structured data: each one answers a different question, and they don't all carry the same weight for your store.

• If Google is what worries you, start with the sitemap and structured data: they're the two Google Search says it uses, and llms.txt changes nothing in its Search.⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}
• If your product pages don't show up in search, the sitemap is a hint, not a fix.⁠^{[9]Google Search Central. Build and submit a sitemap} First find out why they aren't indexed.
• If you want a specific agent to understand your catalog, an llms.txt won't hurt your rankings in Google and is little work if your platform already generates it,⁠^{[8]Google Search Central. Google's Guide to Optimizing for Generative AI Features on Google Search}⁠^{[18]Shopify Dev Docs. llms.txt.liquid} but only your logs will tell you whether that agent requests it.
• If someone offers you llms.txt as a replacement for the sitemap or the markup, ask them which agent reads it and how they're going to check.

Keeping two of these pieces up to date on a catalog that keeps changing is a separate job, and there are tools that do it for you. Clione, for example, generates an llms.txt and a Markdown document for every product, category, collection and page, served on the store's own domain; publishes a JSON-LD graph on each product page with the organization, the website, the breadcrumbs and the product; and then reads the published product page and alerts you when a verified signal disappears. The page on Clione's files for agents has the details.

To keep going, the crawling and indexing reading path puts the terms in this comparison in order, along with the ones behind them.