llms.txt for Law Firms: What the Evidence Says (and Why We Ship It Anyway)
No major AI provider confirms reading llms.txt, and a 137,000-site study found 97% of the files received no bot traffic. Ship it in ten minutes, then spend the real effort on schema, answer-first pages and mentions.
Short answer: llms.txt is a plain-text summary of your site meant for AI engines. As of 2026, no major provider (OpenAI, Anthropic, Google, Perplexity) confirms using it, and a 137,000-site study found that 97% of published files received zero bot traffic. It costs ten minutes and cannot hurt, so we ship one on every site. It is not a strategy, and any agency selling it as one is selling you the cheapest line item they have.
What is llms.txt?
A proposal from September 2024 for a Markdown file at /llms.txt that describes what a site is, what it offers and which pages matter, written for language models rather than search crawlers. The idea is sensible: a model with a context window should get a curated map instead of crawling a menu. Our own file is at isonn.com/llms.txt and reads like a one-page brief: who we are, what we do, our prices, where we work, the twenty pages that matter.
Does any AI engine actually read it?
Not on the record. The state of the evidence in 2026:
| Signal | What it shows |
|---|---|
| Provider documentation | OpenAI, Anthropic, Google and Perplexity document their crawler user agents and robots.txt behaviour. None documents llms.txt. |
| Server logs, 137,000 sites | 97% of llms.txt files received no requests from any known AI bot over the study window. |
| Google's public statements | Google's search advocates have said, more than once, that Google does not use llms.txt and that it resembles the abandoned keywords meta tag. |
| Our own logs, isonn.com | We see retrieval bots fetch practice pages and the sitemap daily. Requests for /llms.txt come from SEO tools and curious humans. |
The crawlers that decide citations fetch HTML pages, read the schema in them, and weigh what other sites say about you. That is where the citation is won or lost.
Why do we ship it anyway?
Three reasons, none of them about ranking:
- The cost is ten minutes. The file is a summary of pages that already exist. If a provider adopts the standard next quarter, sites that have it are ahead by a day.
- It forces a useful exercise. Writing "here are the twenty pages that matter and one sentence on each" exposes sites that have no pages that matter. Firms discover that their practice areas live on one page and their cities on none.
- It is a decent brief for humans. A prospective client, a journalist or a partner at another firm can read it in a minute. We link ours from the footer of nothing, but it is there.
What does move AI citations for a law firm?
The same signals every audit we run comes back to, in the order they usually pay off:
- Crawl access. Retrieval bots must be able to fetch the page. Many firms block them without knowing it; see which AI crawlers to allow.
- Entity consistency. The same firm name, address and phone on the State Bar, Avvo, Martindale, Justia, Yelp and Google. Engines cross-check these before naming anyone.
- Schema that links the firm to those profiles.
LegalServicewithsameAs,Personfor each attorney,FAQPageon question content. See schema markup for law firms in 2026. - Answer-first pages. A question heading and a 40-to-60-word answer at the top; about 55% of cited passages come from the first 30% of a page.
- Mentions and reviews. Ahrefs' 75,000-brand study found unlinked mentions correlate with AI visibility at 0.66 and backlinks at 0.22. Reviews on Google and Avvo are mentions the engines trust most.
Nothing on that list is a text file. Everything on it is work.
How can you tell if an agency is selling you llms.txt as a product?
- It appears as a named deliverable with its own line on the proposal.
- The proposal cannot say which engine reads it or how the effect would be measured.
- "AI optimization" in the proposal consists of llms.txt, a robots.txt edit and "AI-friendly content" with no citation baseline.
A serious AI search engagement starts by measuring your current citation rate across ChatGPT, Perplexity, Gemini and Google AI Overviews, then works the list above. The free AI visibility check is that baseline, and the llms.txt file is item nineteen of twenty-two on the AI citation checklist.
If you want one, here is the template
Keep it under 300 words. Plain Markdown, no marketing language, absolute URLs.
# Sample Family Law, APC
> Family law firm in Glendale, California, serving Glendale, Burbank and the
> Verdugo cities in English and Armenian. Divorce, custody, support and
> domestic violence restraining orders. Not a personal injury firm.
## Practice areas
- https://samplefirm.com/divorce/ : Divorce in Los Angeles County
- https://samplefirm.com/child-custody/ : Custody and visitation
- https://samplefirm.com/glendale-courthouse/ : What happens at the Glendale Courthouse
## Contact
- Phone: +1 (818) 555-0142
- Address: 123 N Brand Blvd, Glendale, CA 91203
- Languages: English, Armenian
Then go work on the schema.
Check your firm
See whether ChatGPT, Perplexity, Gemini and Google AI Overviews name your firm for your practice area and city.
Free AI visibility check