If somebody has told you that your site needs an llms.txt file, it is worth knowing what the evidence actually supports before you spend a morning on it. The short version: it is a proposal rather than an adopted standard, Google has said it does nothing for rankings, and no major AI provider has confirmed reading it in production.
That does not make it useless. It makes it something quite different from how it is usually sold.
What the file is
llms.txt is a plain text file written in Markdown that sits at the root of your site, so it loads at yourdomain.com/llms.txt.
Its purpose is curation rather than permission. Where robots.txt says who may crawl what, and sitemap.xml lists everything you have, llms.txt is meant to say: of all these pages, here are the ones that matter, and here is what each covers.
The reasoning behind it is sound. An AI system answering a question in real time has a limited budget for how much it can read. A short, well-organised map of your best pages is more useful to it than a sitemap listing four hundred URLs with no indication of which matter.
What it looks like
# Digital Marketing Cube
> A digital marketing agency in Delhi offering SEO,
> Google Ads, website development and maintenance.
## Services
- [SEO Services](https://example.com/seo-services/): Technical SEO,
local SEO and content strategy for Indian businesses.
- [Google Ads](https://example.com/paid-advertisement-services/):
Campaign setup, management and account audits.
## Guides
- [WordPress security](https://example.com/blog/post/): How to find
and close the entry point after a repeat infection.That is the whole format. An H1 with the name, a blockquote summary, then H2 sections grouping links with a one-line description each.
What the evidence actually says
This is where the honest answer diverges from most of what is published about it.
Google has said it does nothing for search. Not that it is a weak signal, that it is not used. Any article presenting llms.txt as a ranking factor is describing something the search engine has explicitly denied.
No major provider has confirmed reading it. As of early 2026, OpenAI, Google, Anthropic and Meta have not publicly committed to using the file in their production systems. Some smaller tools and frameworks do support it.
The tooling grew faster than the adoption. There are generators, validators and checkers, and their existence is often presented as evidence the standard is established. It is evidence that people are being told they need one.
A large share of what is written about llms.txt is published by people selling AI visibility services. That does not make it wrong, and it does explain why so much of it assumes the file works and skips the part where the companies who would have to read it have not said they do.
The one use case that clearly holds up
Developer documentation. AI coding assistants read documentation live while helping someone write code. A file that points them at the right pages saves them from reading the entire site, and support here is genuine rather than hoped for.
If you publish an API reference, a plugin manual or technical guides that people ask AI tools about while working, llms.txt is doing a real job today.
If you run a service business, a shop or a local company, that use case does not describe you.
So should you add one?
Our position is straightforward. It cannot hurt, it costs about an hour, and it may become useful if adoption follows. Add one if you have the time going spare. Do not add it instead of the work that actually decides whether AI tools mention you, and do not let anyone bill you for it as an AI visibility strategy.
How to create one properly
- List your genuinely useful pagesTen to thirty, not four hundred. Services, your best guides, your contact and about pages. Curation is the entire point.
- Write one line per pageWhat the page covers and who it is for. Write it for a reader, not as a keyword string.
- Group them under clear headingsServices, Guides, Company. Keep it scannable.
- Open with a two-line summaryWhat the business does and where. This is the part most likely to be quoted.
- Save as llms.txt and upload to the rootSame folder as robots.txt. Confirm it loads at yourdomain.com/llms.txt.
- Update it when your services changeA file listing pages that no longer exist is worse than no file.
Plugins that generate one
Several WordPress plugins now produce llms.txt automatically, and Rank Math has added generation in recent versions. They work by listing your published pages, which produces a file with no curation in it, and curation is the only reason the format exists.
Use a plugin to create the structure if you like, then edit it by hand so it lists what matters rather than everything you have ever published.
What actually gets a business mentioned by AI tools
Since this is what people usually want when they ask about llms.txt, it is worth naming the things that do carry weight.
- Answer the question in the first two sentencesAI systems extract direct answers. A page that opens with three paragraphs of preamble gives them nothing to quote.
- Use question-shaped headingsStructure content the way people ask, and put the answer immediately below the heading.
- Add schema markupOrganisation, LocalBusiness, FAQPage and Article give machines explicit facts instead of leaving them to infer.
- Get mentioned elsewhereBeing cited on sites AI systems already trust does more than anything on your own domain. This is slow and it is the real work.
- Keep your business details consistentName, address, phone and services identical across your site, your Google Business Profile and every directory.
- Publish things only you can writeOriginal data, real numbers from your own work, specific examples. Generic content has nothing to cite.
llms.txt is a reasonable idea waiting for adoption. Adding one is cheap and harmless. Believing it will change whether ChatGPT recommends your business is not supported by anything the companies behind these systems have said.
How to check whether AI tools mention you at all
This takes fifteen minutes and tells you more than any file will.
Write down the ten questions a customer would ask before hiring someone like you. Not your service names, the questions. Then put each one to ChatGPT, Gemini and Perplexity, and note who gets named and which sources are cited.
You will usually find the same handful of sources cited repeatedly. Those are the pages worth being mentioned on, and getting there is a matter of being genuinely useful and genuinely cited, which is the same work good SEO has always been. Where a site is technically slow or badly structured, AI crawlers struggle with it exactly as search crawlers do, and that is a development problem before it is a content one.
Parshant Hadala
Tarun Singla