Quick answer: llms.txt is a proposed standard — a plain markdown file at your site's root (yoursite.com/llms.txt) that gives AI systems a curated, easy-to-read index of your most important content. Whether you need one is 2026's most genuinely contested SEO question: Google's Search team has said on the record that it does not use llms.txt and has no plans to, comparing it to the long-dead keywords meta tag — while parts of the industry recommend it as an emerging best practice. Our honest verdict: it's a 20-minute, zero-risk, low-probability bet — worth adding if you treat it as a lottery ticket, a waste of energy if you treat it as a strategy.
Most coverage of llms.txt picks a side and oversells it. This article lays out what the file actually is, what each side's evidence says, and gives you a decision you can defend.
What is llms.txt, exactly?
llms.txt is a proposal (originated by Jeremy Howard of Answer.AI in late 2024, documented at llmstxt.org) for a standardized file that solves a real problem: AI systems have small context windows and websites are bloated with navigation, scripts, and boilerplate that make extracting the substance hard. The file is written in markdown — human-readable, LLM-friendly — and contains a short description of the site plus a curated list of links to your most important pages, optionally annotated.
Think of it as the difference between handing someone your whole filing cabinet versus a one-page index of what matters. A companion variant, llms-full.txt, goes further and includes the full text of key content in one file.
What llms.txt is not: it's not robots.txt (which controls what crawlers may access — permission), and it's not a sitemap.xml (which lists everything for indexing — inventory). llms.txt is a recommendation — "here's what matters and where to find it." Three different files, three different jobs.
What's the case against it?
The strongest evidence against comes from the biggest player. Google's Search Relations team has stated directly that Google does not support llms.txt — John Mueller publicly compared it to the keywords meta tag, the canonical example of a signal search engines learned to ignore — and Google's guidance states such files "will neither harm nor help" visibility because Google Search ignores them entirely. Since Google's AI surfaces (AI Overviews, AI Mode) draw on Google's index, llms.txt does nothing for the AI surface with the most users.
Beyond Google, the skeptic case has two more legs. First, no major AI engine has publicly committed to honoring it — OpenAI, Anthropic, and Perplexity have not announced that their answer pipelines use llms.txt. Second, practitioner server-log studies through 2025 found AI crawlers fetching llms.txt files inconsistently at best — being fetched occasionally is not the same as being used to shape answers. There is currently no rigorous public evidence that adding llms.txt increases citations or AI visibility.
What's the case for it?
The advocate case is honest when it's framed as a bet, not a result:
The problem it addresses is real. AI systems genuinely do struggle with bloated HTML, and token efficiency genuinely matters to retrieval pipelines. A standard solving a real problem has a plausible adoption path — robots.txt itself started as an informal convention that everyone eventually honored.
Adoption costs nothing and carries no penalty. Even Google's dismissal concedes it "will neither harm nor help." The file takes minutes to create, doesn't interact with your rankings, and can't break anything.
Some AI-adjacent tools and crawlers do read it. Documentation platforms auto-generate it, some developer-focused AI tools consume it, and there are fetch records in server logs. Adoption is real among AI-forward software companies — it's engine commitment that's missing.
If it wins, early adopters win retroactively. Standards flip fast in this space. If any major engine announces support, sites with the file already in place benefit from day one — and everyone else spends that quarter catching up.
So should you add one? Our verdict by situation
If you have 20 spare minutes: yes, add it — with your expectations set correctly. It's a costless hedge on a possible future, not a visibility tactic for the present. Rank it dead last on your GEO to-do list, after the things with actual evidence behind them (crawler access, answer-first content, statistics and citations, structured data, off-site mentions).
If you're choosing where to spend limited effort: skip it without guilt. Every hour spent hand-crafting an llms.txt is an hour not spent restructuring a page opening or getting into a source AI engines already cite — moves with measured effects (the KDD 2024 Princeton study's 30–40% visibility gains from statistics and citations). Google's own advice — put the effort into genuinely crawlable, well-structured content instead — is, on the evidence, correct about priorities even if you disagree about the file.
If someone is charging you money to "implement llms.txt optimization": walk away. A 20-minute markdown file sold as a paid deliverable, on a standard no engine honors, is the kind of certainty-theater this young industry has too much of.
How to add llms.txt in 20 minutes (if you're taking the bet)
Step 1: Create the file
Create a plain text file named llms.txt, in markdown, structured as: an H1 with your site/company name, a one-paragraph blockquote summary of what you do, then H2 sections (e.g., "Products," "Guides," "About") each containing a short list of links with one-line descriptions.
Step 2: Curate ruthlessly
Pick 10–30 of your genuinely important pages, not your whole sitemap. The entire value proposition is curation — a dump defeats the purpose.
Step 3: Upload it to your site root
It should resolve at yoursite.com/llms.txt, alongside robots.txt.
Step 4: Keep it honest and current
Update it when your key pages change; a stale index is worse than none.
Step 5: Optionally add llms-full.txt
If you have cornerstone content worth including in full, the companion file carries the full text. Most SMBs can skip this.
Step 6: Re-check the landscape quarterly
The single fact that would change this article's verdict is a major engine announcing support. That's the trigger to move llms.txt from "lottery ticket" to "required".
The bigger lesson: how to handle contested tactics in a young discipline
llms.txt is a case study in how to think about GEO advice generally. This field is roughly two years old; it has one landmark peer-reviewed study, a fast-moving practitioner consensus, and a strong commercial incentive to oversell every new tactic. The useful filter is a two-question test: What's the cost if this does nothing? and What's the evidence it does something? llms.txt scores near-zero cost, near-zero evidence — hence: fine as a hedge, foolish as a strategy. Plenty of louder tactics (paid "vector optimization," guaranteed "AI rankings") score high cost, zero evidence — those you run from.
That two-question honesty is also how we build SeeGeo: the audit flags as critical only what the evidence actually supports — blocked AI crawlers, content invisible without JavaScript, missing answer structure. llms.txt isn't on that list, and we won't sell you certainty about it in either direction. If a major engine ever commits to the standard, expect this article's verdict to change — in public, on this page.
Frequently asked questions
What is the difference between llms.txt and robots.txt? robots.txt controls permission — which crawlers may access which parts of your site. llms.txt is a recommendation — a curated, markdown-formatted index of your most important content for AI systems that choose to read it. robots.txt is universally honored; llms.txt is a proposal no major engine has committed to.
Does Google use llms.txt? No. Google's Search Relations team has said on the record that Google does not support llms.txt and has no plans to, with John Mueller comparing it to the abandoned keywords meta tag. Google's guidance states the file neither helps nor harms, because Google ignores it.
Does ChatGPT or Perplexity read llms.txt? Neither OpenAI nor Perplexity has publicly committed to using llms.txt in their answer pipelines. Server logs show AI crawlers occasionally fetching the file, but fetching is not the same as using it to shape answers. As of 2026 there is no rigorous public evidence llms.txt increases AI citations.
Can llms.txt hurt my SEO? No. It's invisible to traditional search ranking systems and, per Google's own statement, neither helps nor harms. The only real cost is the time spent creating and maintaining it — and stale, inaccurate llms.txt files are the one way to make it counterproductive if an engine ever does read it.
What should I do instead of (or before) llms.txt? The evidence-backed priority list: ensure AI crawlers aren't blocked (robots.txt and CDN settings), make key content readable without JavaScript, restructure important pages answer-first with statistics and cited sources, add FAQ/HowTo structured data, and build presence in the third-party sources AI engines already cite. Every one of those has stronger evidence than llms.txt; do them first.
How do I create an llms.txt file? Write a markdown file: H1 with your company name, a blockquote paragraph summarizing what you do, then H2 sections listing 10–30 of your most important pages as links with one-line descriptions. Save it as llms.txt at your site root. Twenty minutes, curated rather than complete, updated when your key pages change.