Learn what llms.txt is, why websites are adding it for AI crawlers, how it differs from robots.txt, and whether it actually improves AI visibility and search performance.
Table of Contents
ToggleAdding llms.txt costs almost nothing and carries almost no risk. Whether it actually lifts your visibility in ChatGPT, Claude, Perplexity, or Google's AI results is a different question entirely, and the honest answer right now is: probably not, for most sites. Here is what the file does, what it does not do, and who should actually bother.
llms.txt is a plain Markdown file hosted at yoursite.com/llms.txt that gives large language models a curated summary of your most important content, without the noise of HTML, JavaScript, or ads. Think of it as a recommended reading list, not a set of rules.
The format is intentionally simple: a single Markdown file with headers and a curated list of links, nothing like the JSON-LD blocks used in schema markup.
# Brandella Journal > Digital marketing, SEO, and AI search insights for marketers. ## Docs - [What Is Schema Markup](https://www.brandella.in/what-is-schema-markup-seo-ai-search-guide/): Complete 2026 guide - [What Is Entity SEO](https://www.brandella.in/what-is-entity-seo-who-vs-what-2026/): Knowledge Graph basics ## Optional - [About Us](https://www.brandella.in/about/): Company background
A minimal llms.txt file: plain Markdown, no schema, no directives
Crucially, it is not robots.txt. It cannot block a crawler or restrict access to anything. It is a map, not a lock, and it has nothing to do with the sitemap discovery mechanism we covered in our XML sitemap guide either.
The proposal has moved fast for something with no official backing at all.
Jeremy Howard, co-founder of Answer.AI, first proposes the /llms.txt convention.
Mintlify rolls out llms.txt across every docs site it hosts, including Anthropic and Cursor.
Yoast begins auto-generating llms.txt for WordPress, and Wix rolls it out globally for users.
Roughly one site in ten now carries the file, after eighteen months of industry conversation.
of sites carry an llms.txt file, based on a SE Ranking study of 300,000 domains
of published llms.txt files received zero requests in a large-scale Ahrefs study
direct llms.txt requests out of 500 million+ monitored AI bot visits in one 90-day study
This is where the story splits, and both sides deserve a fair hearing.
According to Ahrefs' analysis of 137,000 domains, 97% of published llms.txt files never got a single request. As reported by Search Engine Land in mid-2025, Google's Gary Illyes stated on the record that Google does not use the file.
John Mueller has compared it to the old, discredited keywords meta tag. According to Derivatex's 2026 llms.txt guide, controlled research found no statistically significant correlation between having the file and receiving more AI citations.
The picture looks different for developer-facing sites. According to SocioLabs' 2026 breakdown, coding agents and IDE tools like Cursor, GitHub Copilot, and Windsurf, along with MCP servers, do actively fetch llms.txt files on documentation-heavy sites. For that specific audience, the file has clear, demonstrable value.
These three files get confused constantly, and the confusion is a big part of why llms.txt gets more credit than the evidence supports.
| File | Purpose | Can It Block Access? | Confirmed Use by Major AI Providers |
|---|---|---|---|
| llms.txt | Curated content summary for LLMs | No | Not officially confirmed by any major provider |
| robots.txt | Crawling instructions | Crawling only, not indexing | Yes, universally respected by compliant crawlers |
| XML sitemap | Discovery aid listing indexable URLs | No | Yes, used by Google, Bing, and most crawlers |
This is where the evidence is strongest. Coding agents and AI-assisted IDEs genuinely fetch and use llms.txt files on docs-heavy sites, making implementation worthwhile.
According to Core dna's 2026 analysis, adding the file costs almost nothing and carries almost no downside, so a low-effort implementation is reasonable even without strong proof of impact.
According to Limy's 2026 llms.txt guide, categories like agentic commerce, vendor research, and comparison shopping are where agent traffic is genuinely heading, making early adoption more defensible.
This is the group most likely to be disappointed. No major AI provider has confirmed using the file as a citation or visibility signal, so treating it as a shortcut around real content and authority work is a mistake.
According to Derivatex's 2026 guide referenced above, a publicly writable llms.txt file is a documented attack surface.
Prompt injection through a tampered llms.txt is a recognized risk in AI security research. Treat the file with the same access controls you already apply to robots.txt and your sitemap.
If you run a documentation site, a developer tool, or an API product, implementing llms.txt is a reasonable half-day project with a real, if narrow, audience already fetching similar files elsewhere.
If you run a general marketing site, blog, or ecommerce store, the more productive use of your time is everything we have already covered in this technical SEO series: clean canonical structure, solid schema markup, and clear entity establishment.
According to LinkBuildingHQ's 2026 analysis, medium and low-traffic sites actually show higher llms.txt adoption than established, authoritative sites, a pattern that suggests smaller sites reaching for an easy win rather than doing the harder, proven work first.
If you do implement it, keep it accurate and current. According to aeo.press's 2026 overview of llms.txt adoption, industry support remains inconsistent even among providers who acknowledge the format, so a stale or inaccurate file is arguably worse than no file at all.
The most common mistake is not writing a bad file, it is writing one and never touching it again. Content changes, pages get removed, and a stale summary file quietly drifts out of sync with the actual site.
A second mistake is confusing llms.txt with a separate, riskier tactic: publishing individual Markdown copies of every page on the site.
If those Markdown files are indexable, they introduce duplicate content at scale, diluting crawl budget and potentially suppressing rankings for the original pages, a problem we covered in more depth in our crawl budget guide.
A third mistake is treating the file's mere presence as a strategy. According to Mintlify's guide to llms.txt, the format works best as a lightweight companion to genuinely well-organized documentation, not as a substitute for organizing that documentation in the first place.
None of this means curiosity about new AI infrastructure is misplaced. Standards that look unofficial today sometimes become genuinely useful within a year or two, and testing low-cost ideas early is a reasonable instinct for any technically minded team.
The mistake is letting a five-minute file implementation substitute for the slower, less glamorous work that actually moves visibility: clear content structure, verified entity signals, accurate schema, and a technically healthy site that both people and machines can navigate without friction.
llms.txt is a curated Markdown summary, not a directive. It cannot block anything.
Only about 10% of sites have adopted it after eighteen months of buzz.
97% of published files receive zero AI crawler requests in large-scale studies.
Coding agents and dev tools genuinely fetch it on documentation-heavy sites.
No major AI provider confirms using it for citations or visibility.
Real SEO fundamentals still matter more than any single text file.










