As of September 2026, our answer is yes, but second. A machine readable pricing file helps an AI agent quote you accurately. It does nothing if your real pricing page is a JavaScript shell or a contact sales form. Fix the page a crawler can read, then publish the file.
This question comes up in almost every project kickoff now. Somebody has read that AI agents are doing product research on behalf of buyers, and they want to know which file to add. It is a reasonable instinct. It is also usually the wrong first move.
So we want to lay out what these files actually are, what the specs say, and where they sit in the order of work.
It is a plain text or Markdown file that lists your plans, prices, limits, and what each tier includes, served at a stable URL like /pricing.md. No layout, no scripts, no interactive toggles. Just the facts an AI system needs to answer a question about what you cost.
The idea borrows from files the web already has. Your robots.txt tells crawlers where they may go. Your sitemap.xml tells them what exists. A pricing file tells them what you charge. Each one trades presentation for parseability.
Nothing about this is standardised yet. There is no governing body for /pricing.md and no search engine that promises to read it. That is worth being clear about before anyone builds a case for it internally.
llms.txt is a proposal from Jeremy Howard, first published on September 3, 2024, with a version 2 update on August 10, 2026. It is explicitly a proposal rather than an official standard, and the specification remains open for community input on its GitHub repository.
The structure is loose. The file can sit at the site root or at any path within it. The only required section is an H1 with the name of the project or site. Beyond that the spec describes an optional byte order mark, a blockquote with a short summary, any number of Markdown sections, and file list sections delimited by H2 headers.
It also asks for something more useful than the file itself. The proposal recommends serving clean Markdown versions of your pages, and using the standard link relations rel="alternate" and rel="describedby" so clients can find them. That second half is the part most implementations skip, and it is the part that would actually change what an agent can read. We cover the format in more depth in our guide to llms.txt.
Yes, and it is the strongest argument in favour. AGENTS.md is a Markdown file that tells coding agents how to work in a repository. It started as a collaboration between OpenAI Codex, Amp, Google Jules, Cursor and Factory, and it is now stewarded by the Agentic AI Foundation under the Linux Foundation.
By its own count it is used by over 60,000 open source projects and supported by more than 20 tools, including GitHub Copilot, Gemini CLI, VS Code, Zed, Aider, Devin, JetBrains Junie and Warp. That is real adoption, achieved in about a year, on a file with no required fields.
The lesson we take from it is that these conventions spread when the tools that read them agree first. AGENTS.md worked because the agent vendors themselves shipped support. No comparable group has agreed to read a pricing file, which is why we would not promise a client any specific return from publishing one.
Because the file is a summary and the page is the source. Every AI system that cites you fetches a URL and reads what comes back. If the pricing on that URL only appears after a script runs, a crawler that does not execute JavaScript sees an empty container, and no side file rescues that.
OpenAI documents three separate crawler tokens for this reason: GPTBot for training, OAI-SearchBot for surfacing sites in ChatGPT search, and ChatGPT-User for actions a person triggers in a chat. Each one is fetching your actual HTML. What they get is what you get quoted on.
We see the same failure repeatedly on client sites we inherit. The pricing looks fine in a browser and is completely absent from the raw response. Nobody catches it because nobody views source on a pricing page. Turning JavaScript off and reloading is a ten second test that finds it every time.
Badly enough that small fixes still move things. In the study Webflow published with its Conf 2026 announcements in September 2026, the median company across 2,000 analyzed websites appeared in only 16% of the AI answers it would want to be part of, and was cited in those answers just 6% of the time.
Meanwhile the traffic maths keeps getting harder. The Pew Research Center reported on July 22, 2025 that when a Google AI summary appeared, users clicked a traditional search result in 8% of visits, against 15% when no summary appeared. Clicks on links inside the summary happened in 1% of visits.
Put those together and the incentive is clear. If fewer people click through, then being quoted accurately inside the answer is worth more than it used to be. A wrong price in an AI answer is now a lost deal you never hear about.
When your pricing is simple, public, and stable. A three tier product with published prices loses nothing by restating them in a text file, and gains a clean canonical version that cannot be misread from a layout. That is a twenty minute job with a real upside.
It is also worth it when your pricing page is unavoidably complex. Usage based pricing with sliders and calculators is genuinely hard for a crawler to interpret. A flat text summary of the same tiers gives the model something unambiguous to quote instead of guessing from a table it half parsed.
Where we would skip it is enterprise pricing that is truly custom. Publishing a file that says contact sales adds nothing an AI could not already tell from your page, and it creates one more artefact to keep in sync.
Two things, and both are fair. First, no major AI platform has committed to reading these files, so the return is speculative. Second, a stale pricing file is worse than none. If your prices change and the file does not, you have published a wrong number in the most machine readable format possible.
We take that second risk seriously. Anything that duplicates a fact on your site needs an owner and a review cadence, or it will drift. In our work the files that survive are the ones generated from the same source as the page, not typed by hand as a one off.
So the honest position is that this is cheap insurance, not a growth lever. Treat it that way and you will not be disappointed by it.
Start with the readable page. Confirm your pricing renders in the raw HTML, that headings describe what each plan is for, and that the numbers appear as text rather than inside an image. Our notes on pricing page design cover the conversion side of the same page.
Then check access. There is no point optimising a page you have quietly blocked. Confirm your robots policy lets the crawlers you want reach it, which we walk through in our piece on controlling AI crawlers in robots.txt.
Only then add the files. Publish llms.txt with an accurate summary and honest links, and a plain pricing file if your pricing is public. Generate both from your CMS if you can, so they update when the page does. That ordering means every step is useful on its own, even if the file conventions never catch on.
Open your pricing page with JavaScript disabled and read what is left. If the prices are missing, that is your entire project for this week, and it matters more than any file you could add. If they are there, you have earned the right to publish the extras.
Most teams find something in that ten second test. If you want another set of eyes on what AI systems can actually read on your site, or help making the pricing page render properly without a rebuild, we are happy to walk through it with you. You can find us at phoenix.studio.
Tell us where you want to go. We'll tell you how we'd get you there.