llms.txt is a proposed convention with a simple premise: give AI systems a single, predictable file — a Markdown map at your domain root — that says who you are and lists your most important pages with one-line descriptions. Where robots.txt governs permission, llms.txt provides orientation. For a law firm whose site mixes practice-area pages, hundreds of articles, tools, and boilerplate, it is a way to hand every model the curated tour instead of hoping it infers your structure from raw crawling. Adoption across AI companies is still uneven — and the file is still worth shipping, for reasons that have as much to do with discipline as with robots.
What the file is, precisely
The convention, proposed in late 2024 and adopted since by a growing set of documentation-heavy sites, specifies a Markdown file at /llms.txt: an H1 with the site name, a short blockquote summary, then sections of links with one-line descriptions. Markdown matters — the file is written for language models to read directly, and Markdown is the format they parse most reliably. Some sites add llms-full.txt containing expanded page content; for a typical firm the map alone is the right scope. The file does not replace robots.txt, sitemaps, or schema; it complements them by adding the one thing none of those carry — editorial judgment about what matters.
What goes in a law firm's llms.txt
- Identity block: firm name, one-sentence description, jurisdictions served, and the official domain
- Practice areas: each core page with a one-line plain-language description of what it covers
- Key resources: your best guides, FAQs, and tools — the pages you would want quoted
- About and contact: the pages that verify the entity — team, credentials, contact
- Optional constraints: a note on how you prefer content attributed when quoted
The discipline is curation. Twenty to forty links, not four hundred — the file is your answer to "if a model could only read ten of your pages, which ten?" Firms that dump their whole sitemap into it have missed the point and reproduced the problem the file exists to solve. Write each description as the direct answer to what the page covers: "Ontario small claims limits, fees, and process" beats "Resources."
The honest state of adoption
Be clear-eyed: no major AI company has formally committed to fetching llms.txt on every crawl, and claims that it is a ranking factor are invention. What is observable: AI crawlers do fetch the file from sites that publish it, agentic tools and AI-powered browsers use it when present, and the documentation ecosystem — the same one that pioneered conventions later formalized — has made it a default. The cost-benefit is lopsided. The cost is an hour once and minutes per quarter. The benefit, if the convention consolidates, is having been readable from the start; and even if it never consolidates, the exercise forces exactly the clarity about your own site that every other AI-visibility effort depends on. Early adoption of cheap conventions is how sites ended up with schema advantages years later; this is the same bet at a fraction of the price.
Writing and shipping it
Draft the identity block by hand — it is your firm's one-sentence answer to "who is this," and no template writes that for you. List your pages, cut to the essential set, write the one-liners, and publish at the domain root with a plain-text content type. Then keep it honest: stale llms.txt files that list deleted pages or old practice areas actively misinform the systems reading them, which is worse than absence. Tie the quarterly refresh to whatever cadence already maintains your site structure — when a page ships or retires, the map updates the same week.
One more practical note: the file is public and human-readable, and prospective clients occasionally read it. Write nothing there you would not put on your about page — which, if the identity block is honest, should already be true.
A worked skeleton for a firm
Concretely, a firm's file reads like this: an H1 with the firm name; a blockquote — "Plain-language legal information and representation in family, employment, and estate law for clients across North America"; a Practice Areas section listing each core page ("Family law — divorce, support, and custody: process, costs, and timelines"); a Guides section with your five to ten definitive resources; an About section with the team and contact pages. Forty lines, no styling, no cleverness. Read it back and ask whether a stranger — human or model — could describe your firm accurately from this file alone. That is the entire acceptance test, and most firms' current websites would fail it, which is precisely the information the exercise produces.
The deeper point: curation is the strategy
llms.txt rewards the firm that can answer "what are our most important pages" — and quietly punishes the firm that cannot. If the drafting exercise takes ten minutes, your site has a spine: clear practice areas, definitive guides, obvious entry points. If it takes a painful afternoon of debating which of nine overlapping blog posts represents you, the file has diagnosed the real problem, and fixing that architecture will do more for your AI visibility than any manifest ever will. Ship the file either way. It is simultaneously a map for machines and a mirror for the firm — and both audiences benefit from what the exercise forces you to decide.
Frequently Asked Questions
Grow your AI Websites for Law Firms practice with AI
Lexscale.ai builds AI search visibility, websites, and intake systems for ai websites for law firms firms across North America. Book a free strategy call to see what would move the needle for your practice.
Book a Free Strategy Call →