llms.txt is a plain text file some websites now publish at their root domain, meant to give AI systems a curated map of a site’s most important pages. It’s been marketed heavily as an essential piece of AI search optimization for law firms and every other business trying to get cited by tools like ChatGPT and Perplexity. Here’s the honest version: current evidence shows it doesn’t meaningfully improve AI citations yet, and firms chasing it instead of the fundamentals that actually work are solving the wrong problem.
What Is llms.txt?
llms.txt is a Markdown file, proposed by developer Jeremy Howard in September 2024, designed to sit at a site’s root address, similar to how robots.txt and sitemap.xml already work. The idea was straightforward: give AI systems a clean, structured summary of a site’s key pages, rather than forcing them to parse a full HTML page cluttered with navigation menus, cookie banners, and scripts.
A typical llms.txt includes a brand name as a heading, a short summary of what the site or business does, and a list of its most important pages with one-line descriptions each. In concept, it’s a shortcut, a cheat sheet handed directly to an AI system instead of making it work to figure out a site’s structure on its own.
What llms.txt Does Not Do
This is the part most marketing around llms.txt leaves out, and it matters directly for any firm being sold this as a priority investment.
Google has been explicit and on the record that it does not use llms.txt for Search, AI Overviews, or AI Mode. Google’s own search advocates have compared it to the old keywords meta tag, a signal that technically exists but that Google simply ignores. Beyond Google, large-scale crawler analysis from Ahrefs, studying over 137,000 domains, found that the retrieval crawlers actually responsible for feeding AI search answers barely request the file at all. GPTBot, ClaudeBot, and PerplexityBot largely skip it in favor of crawling standard HTML directly.
Adoption numbers tell a similar story. Independent studies put llms.txt adoption somewhere between 6% and 10% of sites analyzed, concentrated heavily in developer documentation and SaaS platforms rather than general business websites. After roughly two years of industry conversation, there’s still no publicly confirmed commitment from any major AI provider, including OpenAI, Google, or Meta, to use llms.txt as a real signal in production search or answer systems.
There is one meaningful exception worth noting. Anthropic has publicly confirmed support for the format, and Perplexity has stated it retrieves llms.txt to help prioritize which pages to consider. That’s a real signal, just not yet a widespread one across the platforms most law firms care about being cited in.
So Why Add It at All?
Given all of that, it’s fair to ask why this file is worth mentioning as part of AI search optimization for law firms rather than dismissing it outright. A few honest reasons:
- The cost is close to nothing. Creating a basic llms.txt takes well under an hour and requires no ongoing maintenance burden beyond periodic updates.
- It’s a hedge, not a strategy. Support from Anthropic and interest from Perplexity suggest the standard could gain broader traction over time. Having one in place costs little now and avoids playing catch-up later if adoption grows.
- It causes no harm when done correctly. A standard llms.txt file poses no downside, provided a firm avoids publishing a companion llms-full.txt, which hands over full page content in an easily scraped format. For most law firms, that tradeoff isn’t worth it.
The honest framing is that llms.txt is a low-cost, speculative bet on future adoption, not a proven lever for AI search optimization for law firms today. Any agency presenting it as a guaranteed way to improve AI citations is overstating what the current evidence actually supports.
What Actually Drives AI Search Optimization for Law Firms Right Now
The mechanisms with real, current evidence behind them look nothing like a single text file. AI systems primarily determine what to cite based on:
- Entity strength and structured data, meaning proper schema markup connecting a firm, its attorneys, and its content into a consistent, verifiable entity, which AI systems can confirm rather than infer
- Answer-first, comprehensive content that directly addresses specific questions clearly, rather than requiring an AI system to dig for a usable answer
- Third-party validation, including consistent information across legal directories, review platforms, and independent mentions, since AI systems weigh outside confirmation more heavily than a firm’s own claims about itself
- Proper indexing across the right search infrastructure, since tools like ChatGPT search draw from Bing’s index while Gemini and AI Overviews draw from Google’s, meaning a site optimized only for one may be invisible to systems relying on the other
None of these require a trendy new file. They require the same disciplined technical and content work that’s always mattered, adapted for how AI systems specifically evaluate and select sources.
How to Actually Build One, If You Want One
For firms that want the low-cost hedge anyway, the format is simple enough to implement without technical help:
- Create a Markdown file with the firm’s name as a top-level heading, followed by a short blockquote summarizing what the firm does
- Add a handful of sections listing the site’s most important pages, typically 10 to 40, each with a one-sentence description of what that page covers
- Publish it at the site’s root address, so it’s accessible at yoursite.com/llms.txt
- Skip publishing a companion llms-full.txt, which exposes full page content for easy scraping
- Revisit and update it periodically, since a stale file pointing to outdated pages provides no benefit even in the unlikely event a system does check it
Frequently Asked Questions
Does adding llms.txt improve a law firm’s chances of being cited by ChatGPT or Perplexity?
Current evidence doesn’t support that. Major retrieval crawlers rarely request the file, and no major AI provider besides Anthropic has confirmed using it as a citation signal. Perplexity has stated it retrieves the file to help prioritize pages, but that’s not the same as a guaranteed citation benefit.
Is llms.txt the same thing as robots.txt?
No. Robots.txt tells crawlers what they’re allowed or not allowed to access, and major AI crawlers document actual compliance with it. llms.txt is a curated content summary with no equivalent enforcement or confirmed usage across most major AI systems.
Should a law firm prioritize llms.txt over other AI search optimization work?
No. Schema markup, comprehensive answer-first content, and consistent third-party validation across directories and review platforms all have stronger, more current evidence behind them. llms.txt is reasonable to add as a low-cost, speculative addition, but it shouldn’t come at the expense of those higher-impact priorities.
