llms.txt in Webflow: What It Is and Whether It Does Anything

An llms.txt file is a short Markdown file at a website's root that hands AI agents a curated map of the site instead of making them crawl it. Webflow has supported uploading one since 23 July 2025, under Site settings > SEO, and the upload is a short documented procedure. On the evidence available today it earns you very little: Google Search states plainly that it ignores the file, and the largest published measurement found no relationship between having one and being cited by AI models. Five things are worth knowing before you add it to a client site.

• Google Search ignores llms.txt. Google's own AI optimization guide names the file and says creating one will "neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."
• Chrome audits for it anyway. Lighthouse ships an llms.txt discoverability audit inside its Agentic Browsing category, which is the same company reaching a different conclusion.
• Adoption sits at 10.13% across roughly 300,000 domains measured by SE Ranking, and the largest sites use it slightly less than mid-sized ones.
• Webflow's implementation is a file upload, capped at 100 KB, UTF-8 only, custom domain only, and it never appears on the .webflow.io staging domain.
• The serious publishers are AI labs documenting their own APIs, per the llms.txt specification, not marketing sites chasing citations.

What is an llms.txt file?

An llms.txt file is a Markdown document, published at /llms.txt, that describes what a site is and links to the pages an AI agent should read first. Jeremy Howard proposed it on 3 September 2024, and the specification at llmstxt.org now carries a version 2 and was last modified on 10 August 2026. The format is deliberately minimal: the only required section is an H1 carrying the name of the project or site. Everything else is optional, including the blockquote summary and the H2 sections that hold lists of links.

The specification chose Markdown over XML on purpose, because the file is meant to be read by language models rather than parsers. It can sit at the root or at any subpath, so /docs/llms.txt covers everything under /docs/. Version 2 goes further and recommends publishing a clean .md version of each page, discoverable through rel="alternate" type="text/markdown" link relations.

What I tell agency clients is that llms.txt is a proposal, not a standard. Nobody ratified it, no browser or search engine is obliged to honour it, and the author himself calls it a proposal on the spec page. If you are deciding whether to treat it as infrastructure or as an experiment, treat it as an experiment.

Does Google use llms.txt?

Google Search does not use llms.txt, and Google has said so in its own documentation rather than in a conference remark. The Google Search guide to optimizing for generative AI features lists the file under a section headed "what you don't need to do" and states: "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." The same page adds that maintaining one "will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."

Google's wording matters for anyone selling AEO work. Google is not saying the file is risky or penalised, and it is not saying it might start reading it soon. It is saying the file is inert for Google Search, including AI Overviews and AI Mode, both of which are grounded in Google's core Search index. When a prospect forwards me an audit that lists a missing llms.txt as a critical AI-visibility issue, that single line is what I send back.

Organisation	Position on llms.txt	In their own words	What that position applies to
Google Search	Does not use the file	"Google Search itself doesn't use them"	AI Overviews, AI Mode and ordinary Search results
Chrome, via Lighthouse	Audits for it, does not require it	A missing file "is marked as Not Applicable (N/A), as providing the file is optional at the moment"	Browser agents driving a page on a user's behalf
Webflow	Ships the upload, hedges the value	"its effects on AI-driven search are unclear"	Any Webflow site published on a custom domain
llmstxt.org, Jeremy Howard	A proposal aimed at agents, not at search	"A proposal to standardise on using an /llms.txt file to provide information to help agents use a website"	Agents fetching information at inference time

Sources, in row order: Google Search's generative AI optimization guide, the Lighthouse llms.txt audit, Webflow's llms.txt help article and the llms.txt specification.

Why does Chrome audit for a file Google Search ignores?

Chrome's Lighthouse tool does audit for llms.txt, in a category called Agentic Browsing, and this is the genuinely confusing part of the picture. The Lighthouse llms.txt audit documentation, last updated 5 May 2026, places the check under "Discoverability" and explains that Lighthouse "flags the pages if a server error occurs when attempting to retrieve the llms.txt file." A 404 is not a failure: the audit is marked Not Applicable, "as providing the file is optional at the moment."

Two different teams inside the same company are therefore pointing in two directions, because they are solving two different problems. Google Search is describing how its ranking and grounding systems work. Chrome is describing how a browser agent, driving a page on a user's behalf, might orient itself faster. The rule I apply when a client raises this contradiction is to separate the two audiences: if the goal is AI Overviews or AI Mode, llms.txt does nothing, and if the goal is browser agents navigating the site, it may help marginally and costs nothing to provide.

How do you add an llms.txt file to a Webflow site?

Adding an llms.txt file to a Webflow site is a file upload, not a page you build or a CMS field you bind. Webflow's help documentation gives four steps, all of them inside Site settings, and once the site is published the file is served at the root of the custom domain, for example https://example.com/llms.txt.

• Open Site settings > SEO > LLMs.txt.
• Click Upload file.
• Choose a file that is UTF-8 encoded and under 100 KB.
• Click Save changes, then publish the site to the custom domain.

One prerequisite comes before those steps, and it is worth stating carefully because Webflow's own materials are inconsistent about it. Webflow serves the llms.txt file only at the root of a custom domain, and the free Starter Site plan ships a .webflow.io address rather than a custom domain, so a site on Starter cannot serve the file at all. That much follows from two documented facts. Which paid tier is sufficient is a different question, and Webflow has published more than one answer to it, so the honest instruction is to confirm the requirement in Site settings on the actual client site before quoting the work rather than to trust any tier list, including this one.

Three further constraints are worth writing into a handoff document. The file is not published to the .webflow.io staging domain, so it cannot be checked before the custom domain goes live. Webflow keeps it out of the search index. And Webflow's implementation is scoped to the file alone: the spec proposes .html.md variants of every page, and Webflow states it deliberately does not generate them. On the builds I run, that last point is what stops the conversation, because half the value in the version 2 spec lives in those Markdown page variants.

How does llms.txt compare to robots.txt and sitemap.xml?

robots.txt, sitemap.xml and llms.txt answer three different questions, and none of them substitutes for another. robots.txt controls access, sitemap.xml lists indexable URLs for search engines, and llms.txt offers a curated summary aimed at agents at the moment they need it. The table below is the comparison I use when an agency asks which of the three is actually worth an hour of a developer's time on a Webflow project.

File	What it does	Where it is set in Webflow	Does Google Search use it?	Format
robots.txt	Tells crawlers which paths they may request	Site settings, editable rules	Yes	Plain text directives
sitemap.xml	Lists indexable URLs for search engines	Auto-generated, static pages toggle	Yes	XML
llms.txt	Gives agents a curated summary and entry links	Site settings > SEO > LLMs.txt, file upload, 100 KB cap	No, stated explicitly	Markdown

If you have one hour and a choice between schema markup and llms.txt on a Webflow site, spend it on schema markup. Schema markup feeds a system Google confirms it uses for rich results, and how to set it up in Webflow is covered in the guide to Webflow schema markup. The robots.txt side of the same decision is covered in the Webflow SEO guide.

How many websites actually publish an llms.txt file?

Adoption of llms.txt reached 10.13% in the largest published measurement, which is a long way from the near-universal presence of robots.txt. SE Ranking's study of roughly 300,000 domains, published on 7 November 2025, found that 10.13% of them served the file. Broken out by traffic band, low-traffic sites of 0 to 100 monthly visits were at 9.88%, mid-traffic sites of 1,001 to 5,000 visits at 10.54%, and high-traffic sites above 100,001 visits at 8.27%.

The shape of that distribution is more interesting than the headline. Calculated from SE Ranking's published percentages, roughly nine domains in ten had no llms.txt file at all, and mid-sized sites led the largest sites by 2.27 percentage points. Both of those are my arithmetic on their figures, not numbers they published. If a new convention were quietly working, the biggest and best-resourced sites would normally be furthest ahead, and here they are marginally behind.

Does llms.txt increase AI citations?

No published evidence shows llms.txt increasing how often a domain is cited by AI models. SE Ranking tested this directly, using Spearman correlation, an XGBoost regression and SHAP analysis over domain-level citation counts. The result they report is blunt: "when we removed the LLMs.txt factor, the model's predictions actually improved. It became more accurate and reliable on test data." A variable whose removal improves a model is contributing noise, not signal.

One study is not the last word, and SE Ranking says as much, noting the findings are context-dependent because the factors interact. But the SE Ranking study is the only measurement at that scale I could find, and it points the same way as Google's own documentation. When I audit a Webflow site for answer-engine visibility, llms.txt is not on the list of things I check first, and the reasoning behind that ordering is in the explainer on answer engine optimization.

Who actually reads llms.txt today?

The heaviest real-world use of llms.txt is software documentation read by coding agents, not marketing sites read by chatbots. The spec page notes that the AI labs publish llms.txt files for their own developer documentation, naming OpenAI, Anthropic and Gemini, and that several documentation platforms and CMSs generate the file automatically for the sites they host.

• Mintlify generates llms.txt and Markdown page versions for every site it hosts.
• GitBook serves an llms.txt file for published documentation sites.
• Yoast SEO generates and maintains an llms.txt file on WordPress.
• AIOSEO ships an llms.txt generator for WordPress.
• Wix generates an llms.txt file for every site on the platform.
• Webflow generates nothing: the file is a manual upload you replace by hand.

The documentation platforms and AI labs above describe the actual constituency for llms.txt. An agent fetching a library's API reference benefits from a curated index because the alternative is crawling a hundred pages of documentation. A chat assistant answering "who builds Webflow sites for agencies" is grounding on a search index instead, which is why the file does nothing for that query. What I see agencies get wrong is pitching llms.txt to a services client as an AI-visibility feature when its demonstrated audience is developer documentation.

What goes wrong with llms.txt on Webflow sites?

The most common llms.txt mistake on a Webflow site is testing the file on the .webflow.io staging domain, where Webflow never publishes it, and three other failure modes account for nearly everything else I run into. Each has a specific consequence rather than a vague one, so they are worth naming individually before a build goes out the door.

• Testing on the staging domain. Webflow does not publish the file to .webflow.io, so a developer checking yoursite.webflow.io/llms.txt sees a 404 and concludes the upload failed. Consequence: the file gets re-uploaded repeatedly or abandoned as broken.
• Pasting marketing copy into the file. Webflow's own guidance warns against pasting homepage hero text verbatim and recommends describing products plainly. Consequence: a model reading the file learns the tagline and nothing checkable about the business.
• Letting it go stale after launch. Nothing in Webflow regenerates the file when pages change, because it is a static upload with no CMS binding. Consequence: within a redesign cycle the file lists URLs that now 404, which is worse than having no file.
• Billing it as an AEO deliverable. Selling llms.txt as the reason a client will appear in AI answers is a promise the published evidence does not support. Consequence: an awkward conversation at the first quarterly review when nothing moved.

Should you add llms.txt to a client's Webflow site?

Add an llms.txt file if the four-step upload fits your build process and someone will commit to updating it, and skip it if either condition fails. The file is small, carries no technical risk, and Google has explicitly confirmed it will not hurt anything. The case for adding an llms.txt file is optionality: the convention has been shipping in Webflow for roughly thirteen months as of publication, calculated from the 23 July 2025 launch date, and if agents converge on it later the groundwork is already done.

The case against is honesty about what it buys. On client projects I add llms.txt when the site already has clean schema markup, a working sitemap and pages that answer real questions, and I leave it off when those are missing, because it would be the fourth-best use of the same hour. The decision rule I use is simple: if llms.txt would be the most advanced thing about the site's machine readability, the site has a bigger problem than a missing file.

Want a second opinion on your Webflow site before you spend on AEO?

If you are weighing llms.txt against schema markup, sitemap structure or a content rebuild and want an outside read on which is worth funding, book a 15-minute call. You get a straight ranking of the three or four changes that would actually move your site's machine readability, in order, with the ones that are not worth doing named as such. No audit document, no deck, just the ordering.