llms.txt Explained, What It Is and Whether It Does Anything
llms.txt is a proposed standard for telling AI models what your site contains. Here is the honest position, what it is, who supports it, and why we publish one anyway.
By HandsOnTech Team · July 23, 2026 · AEO, llms.txt, AI search, technical SEO
llms.txt is a proposed convention, introduced in 2024, for publishing a Markdown file at your domain root that lists and describes your site’s important pages for AI models. It is not an official standard and no major model provider has confirmed using it. It costs about an hour to produce and has no known downside.
Let us be direct, because most writing on this topic is not: there is no public evidence that llms.txt affects whether you get cited by any AI system. It is a proposal that gained traction on the strength of being a sensible idea, not on the strength of adoption by the companies that would need to consume it.
We publish one. Here is the reasoning, and the honest caveats.
What is llms.txt, exactly?
A Markdown file at https://yourdomain.com/llms.txt. The convention is:
- An H1 with your organization name
- A blockquote with a one-paragraph description
- Key facts, location, contact details
- H2 sections grouping your pages, each entry a URL followed by a short description
The idea is that a model with a limited context window, asked about your business, could read one compact file describing your site rather than crawling and reading hundreds of pages. As a design intuition that is reasonable. It is also the same intuition behind a well-structured sitemap and clear page titles, which already exist and are actually consumed.
The honest status
| Question | Answer |
|---|---|
| Is it a standard? | No: a community proposal from 2024 |
| Do OpenAI, Anthropic, Google or Perplexity use it? | None has confirmed doing so |
| Does it affect rankings? | No evidence |
| Does it affect citations? | No evidence |
| Can it hurt? | Not if it is accurate |
| Time to produce | About an hour for a mid-sized site |
The cost-benefit is unusual: a plausible but unproven benefit against an almost-zero cost. That is enough to justify doing it, and not enough to justify anyone charging you a project fee for it.
Why publish one anyway?
Two reasons, and the second is the real one.
Cheap option value. If adoption arrives, the file is already there. If it never does, an hour was spent.
The audit is the product. Writing an accurate one-line description of every page you want found is a forcing function. You have to decide which pages are worth listing, which means confronting the ones that are not, near-duplicates, thin templated pages, orphans with no internal links, pages you forgot existed.
We have never produced one of these files for a client without finding at least one page that should not exist and one that should have been linked from the navigation. That finding is worth considerably more than the file.
How to write one that is useful
-
List only pages you want found. Not admin routes, not paginated archives, not tag pages, not the three near-identical service pages you have been meaning to consolidate. The exclusions are the informative part.
-
Describe, do not just link. A bare URL adds nothing over your sitemap.
- https://example.com/services/staff-augmentation/: Senior engineers embedded in your team, monthly contract, U.S. hoursis a sentence that carries information. -
Group by what a buyer would ask for. Services, locations, case studies, tools, blog. Not by your CMS’s content types.
-
Put your entity facts at the top. Legal name, address, phone, email. If a model reads only the first ten lines, those are the ten lines that should be there.
-
Keep it accurate. A file listing URLs that 404 or redirect is worse than no file. Check it whenever you change your URL structure, this is the failure mode we see most often, because nothing warns you when it goes stale.
What to do instead if you only have an hour
If you must choose between publishing an llms.txt and fixing your robots.txt, fix robots.txt. That file genuinely determines whether GPTBot, ClaudeBot, PerplexityBot and Google-Extended can read your site at all, and a wrong line in it makes every other AEO effort pointless.
After that, the order is: make sure commercial pages are indexed, serve real server-rendered HTML, add direct answer blocks to money pages, make your entity data consistent. llms.txt belongs somewhere below all of those.
Anyone selling llms.txt as a primary AEO tactic has the priority order backwards.
Related: AEO vs SEO, how to get mentioned in ChatGPT, and our AEO overview. Ours is public at /llms.txt if you want a working example. For a technical audit, see SEO & AEO services or contact us.
Got questions?
Is llms.txt an official standard?
No. It is a community proposal introduced in 2024, not a specification from any standards body and not something the major model providers have committed to consuming. It is a convention that some sites adopt, nothing more.
Does llms.txt improve rankings or citations?
There is no evidence that it does. No major provider has confirmed using it as a signal. Anyone presenting it as a ranking factor is overstating what is known.
Then why publish one?
It costs about an hour, cannot hurt, and forces a genuinely useful exercise, writing one accurate sentence about every page you want found. That audit surfaces orphaned and duplicate pages more often than people expect. Treat the byproduct as the real value.
How is llms.txt different from robots.txt?
robots.txt controls access and is universally honoured by legitimate crawlers. llms.txt describes content and is honoured by nobody in particular. If you only do one thing, get robots.txt right, that one genuinely determines whether AI crawlers can read your site.
Where does the file go and what format is it?
At your domain root, as /llms.txt, in plain Markdown. Convention is an H1 with your organization name, a blockquote summary, then H2 sections containing bulleted links with a short description after each URL.
Should I list every page?
List every page you want found and nothing else. Excluding admin routes, thin pages and duplicates is the point. A file listing 800 URLs with no descriptions provides no more information than your sitemap already does.
Built for founders who are done waiting on agencies.
One team. One contract. Working software every two weeks.
Tell us what you're building.
You'll get a plan, a price and a date, usually within two business days. No sales deck, no discovery-call maze.
- ✉️sales@handsontech.ioWe reply within one business day
- 📞(866) 965-8749Mon–Fri, 9am–6pm ET
- 📍2108 N ST STE NSacramento, CA 95816
- ⚡Free website auditRecorded walkthrough back in 48 hours
- 🔒NDA on requestSigned before you share anything sensitive