Does Law Firms Get Recommended by AI?
Custom Legal Marketing asked four AI platforms to recommend lawyers 12,800 times and then visited every law firm website they named to see whether the firms being recommended actually had either one.
Hey AI, do you care about llms.txt?
Hey AI, does this matter?
Does llms.txt Actually Help Law Firms Get Recommended by AI?
Law firms are being told that AI assistants need a special file, or a footer link that says “Hey AI, learn about us,” before they will recommend a lawyer. Custom Legal Marketing asked four AI platforms to recommend lawyers 12,800 times and then visited every law firm website they named to see whether the firms being recommended actually had either one.
The short version: 63.7% of the law firms that ChatGPT, Claude, Gemini and Grok recommended have no llms.txt file at all. Only 3.8% have an AI-directed link on the page. Firms with the file did not land higher in AI answers than firms without it, and the firms that all four platforms agreed on were the least likely to have one.
The Study on llms.txt at a Glance
Four AI platforms, 3,200 lawyer recommendation prompts each, and a visit to every law firm website they returned.
Prompts run November 9, 2026 across ChatGPT, Claude, Gemini and Grok. Websites visited November 10, 2026.
What Law Firms Are Being Told About llms.txt
The llms.txt file started as a proposal published in September 2024 by technologist Jeremy Howard. The idea is a plain text file at the root of a website that gives AI tools a short, clean summary of the site and links to its most important pages. The proposal describes its purpose as helping AI agents use a website.
The idea has since been picked up by the marketing world. In June 2025 the WordPress plugin Yoast SEO added a setting that generates the file automatically, and files carrying the signatures of other SEO plugins appear throughout this study. A visible footer link has also started appearing on law firm websites, usually worded “Hey AI, learn about us,” that points AI crawlers to the file or to a dedicated page written for them.
The companies that build the AI systems have been less enthusiastic. Google’s John Mueller said in April 2025 that none of the AI services had said they use the file and that server logs show they do not check for it. He called it “comparable to the keywords meta tag,” a self-description that search engines stopped trusting years ago.
That leaves law firms with a sales pitch on one side and a shrug on the other. This study looks at the question from a third direction: the websites of the firms AI platforms actually recommend.
Google’s John Mueller called the llms.txt file “comparable to the keywords meta tag.”
How the Study Worked
Powered by CLM Sequoia
CLM Sequoia sent the same style of request, “recommend a [practice area] lawyer in [city],” to ChatGPT, Claude, Gemini and Grok. The prompts covered 30 practice-area terms in 100 U.S. cities, which comes to 3,200 prompts per platform and 12,800 in total. The platforms returned 55,512 website citations pointing to 7,502 different domains.
What Was and Wasn’t Crawled
The study is about law firm websites, so each address the platforms returned was checked for an llms.txt file, for links aimed at AI systems, and for the wording of those links only if it led to a working law firm website. The analysis covers the 3,212 distinct law firm websites that met that test. The following were left out:
- Directories and lawyer referral platforms, a group covered in an earlier report on the legal directories AI platforms lean on
- Bar associations and other professional associations
- Courts and government agencies
- Legal aid groups and public defender offices
- Media and general reference sites
- Nonworking addresses that did not connect to any website
- Parked domains, domains listed for sale, and placeholder or under-construction pages
- Addresses occupied by a business unrelated to law
- Inaccessible websites that refused automated visits or returned an error
Addresses that redirect to the same firm were counted once.
Most Recommended Firms Have No llms.txt File
Of the 3,212 law firm websites, 1,146 (35.7%) had an llms.txt file with real content in it. Another 21 returned a file at that address that held no heading and no links, and those are counted separately. The remaining 2,045 firms (63.7%) had nothing there. Every one of them was recommended by at least one AI platform anyway.
Half of the files that do exist were not written by anyone. Of the 1,146 files, 574 carried the signature of an SEO plugin that builds the file automatically from the site’s existing pages. On WordPress sites 39.1% had the file, compared with 26.1% of sites built on anything else. For those firms the file may reflect nothing more than a plugin setting that is switched on.
llms.txt on AI-Recommended Law Firm Websites
Share of 3,212 recommended law firm websites, by what was found at the llms.txt address.
Which WordPress Plugins Are Writing These Files
The plugin-built files come from a short list of tools. Of the 574 llms.txt files that carried a plugin signature, 67.6% were generated by Yoast SEO, 17.2% by Rank Math and 14.3% by All in One SEO. Plugins built only to create llms.txt files accounted for the remaining 0.9%. The first three are general SEO plugins, and the llms.txt file is one feature among many in each of them.
The WordPress Plugins Behind the Files
Share of plugin-generated llms.txt files on recommended law firm websites, by the plugin that built them.
Based on the llms.txt files that carried an identifiable plugin signature. Files with no signature are not included.
Adoption also tracks practice area closely. Among firms recommended mainly for injury work, 48.2% had the file. Among firms recommended mainly for business and corporate work, 17.9% did. That uneven spread matters when reading the comparison further down.
Is the “Hey AI” Link Helping or Virtue Signaling?
The visible link is much rarer than the file. Only 122 of the 3,212 firms (3.8%) had a link on the page aimed at AI systems. In 117 cases it sat in the footer, three firms placed it in the header navigation, and two put it in the body of the page.
The wording is remarkably uniform. Seventy-five of the 122 firms used some version of “Hey AI, learn about us,” and 58 of those used exactly those words. The second most common label was “AI Info” or “AI Information.”
What the AI Link Says
Anchor text on the 122 recommended law firm websites that link to an AI page or file.
| Link text | Share of firms with an AI link | Share |
|---|---|---|
| “Hey AI, learn about us”Includes versions that name the firm | 61.5% | |
| “AI Info” / “AI Information” | 18.0% | |
| “LLMs.txt” / “LLM Info” | 8.2% | |
| “AI & LLM Info” | 3.3% | |
| “For AI” | 2.5% | |
| Other“AI Factsheet,” “AI Profile,” “AI Entity,” “AI Bot Learning Guide” and similar | 6.6% |
Where the link goes is just as templated as what it says. Sixty-eight firms link to a dedicated page written for AI systems, and 40 of those pages live at the same address on each firm’s site, /llm-info/. Fifty-two firms link straight to the llms.txt file. Nineteen of the 122 firms had the link without having an llms.txt file at all.
Identical wording at an identical address across dozens of unrelated law firms is what a vendor template looks like. Among the 35 link firms whose footer displayed a website credit, at least 17 credited the same marketing agency. That shows how the tactic spreads. As for whether it helps, the comparison in the next section found that firms with the link held essentially the same positions in AI answers as everyone else.
Does Any of It Change Who Gets Recommended?
This is the part law firms are paying for, so each measure is laid out separately. The comparison is between the 1,146 firms with an llms.txt file and the 2,045 firms with none.
Firms With the File vs. Firms Without It
How often, how broadly and how high each group was recommended across 12,800 prompts.
| Measure | With llms.txt 1,146 firms | Without 2,045 firms | Result |
|---|---|---|---|
| Average position in the AI answer | 2.62 | 2.61 | No difference |
| Average best position achieved | 2.02 | 2.04 | No difference |
| Average number of platforms naming the firm | 1.35 | 1.37 | No difference |
| Median times recommended | 2 | 2 | No difference |
| Average times recommended | 5.04 | 4.19 | Small difference |
| Recommended only once | 34.2% | 41.4% | Small difference |
Position 1 is the first firm named in an answer. The 21 firms with an empty placeholder file are excluded from both columns.
Position did not move. When an AI platform lists several firms, the order matters, and firms with the file averaged position 2.62 against 2.61 for firms without it. The most direct test is inside a single answer. In 2,388 answers that named at least one firm from each group, the firms with the file held the better average position 43.0% of the time, the worse position 44.8% of the time, and tied in the rest. Of all the first-place recommendations in the study, 40.6% went to firms with the file, almost exactly their 40.2% share of recommendations in every position.
Agreement across platforms did not move either. A firm named by ChatGPT, Claude, Gemini and Grok alike has cleared four different systems. As an earlier CLM study found, AI platforms rarely recommend the same law firms, so that group is small: 93 firms. Only 23 of them (24.7%) had an llms.txt file and three had an AI link. The firms every platform agreed on were the least likely to have the file.
Inside the Same Answer, and Across Platforms
Two tests of whether the file gives a firm an edge over the firms recommended beside it.
Compares the average position of each group within one platform’s answer to one prompt.
Number of AI platforms that recommended the firm at least once.
The number of recommendations showed a small gap, and it deserves a straight reading. Firms with the file were recommended 5.04 times on average and firms without it 4.19 times. The typical firm in both groups was recommended twice. The relationship is faint: knowing whether a firm has the file explains less than half of one percent of the variation in how often it was recommended.
Two things in the data caution against crediting the file for that gap. The first is the mix of firms. Injury firms are the most likely to have the file, and 11 of the 30 practice-area terms in the study were injury terms, so injury firms had more chances to be named. Sixty-two of the 100 most recommended firms in the study are injury firms. The second is what happens when firms are compared only with others in their own practice group: the gap holds up in two of the six groups and disappears in the other four.
The Gap by Practice Area
Average number of AI recommendations for firms with and without an llms.txt file, grouped by the practice area each firm was recommended for most.
| Practice group | Firms | Have llms.txt | Avg. recommendations with file | Avg. recommendations without | Result |
|---|---|---|---|---|---|
| Injury | 996 | 48.2% | 6.49 | 5.28 | Small difference |
| Family | 346 | 39.0% | 3.99 | 3.08 | No reliable difference |
| Criminal defense | 508 | 38.4% | 3.64 | 2.42 | Small difference |
| Estate and probate | 401 | 32.2% | 2.83 | 2.60 | No reliable difference |
| Employment | 342 | 29.2% | 4.60 | 3.64 | No reliable difference |
| Business and corporate | 598 | 17.9% | 5.53 | 5.80 | No reliable difference |
The footer link tells the same story on a smaller scale. The 122 firms with an AI link averaged 4.84 recommendations against 4.48 for everyone else, with the same median of two and nearly identical positions. Injury firms with the link did get recommended more often than injury firms without it, 7.17 times against 5.77. That is 63 firms, and they are firms that have hired a vendor to work on AI visibility. The study has no way to separate the link from the rest of the work being done for them.
The fairest summary of the evidence is this. Nothing in the data shows the file or the link moving a firm up within an answer or getting it named by more platforms. There is a modest difference in how often firms are named, and it is confined to two practice areas. CLM’s reading of that pattern is that the file marks a more actively marketed firm. A signal that AI platforms act on would be expected to show up in every practice area and in position as well.
What This Means for Law Firms
An llms.txt file is not required to be recommended. Nearly two out of three recommended firms have none, including three out of four of the firms all four platforms agreed on. A firm that has been told it is invisible to AI without one can check that claim against 2,045 firms that were recommended anyway.
The file is cheap, and that is the right price. A plugin can generate it with one setting, which is how half of the files in this study came to exist. There is no evidence here that having one causes harm. A firm paying a recurring fee for it, or being told it is the centerpiece of an AI strategy, is paying for something this data does not support.
Think twice about the footer link. A line reading “Hey AI, learn about us” is visible to every prospective client who scrolls to the bottom of the page, and 52 firms use it to send those visitors to a raw text file. It was found on 3.8% of recommended firms, and their positions in AI answers were essentially the same as everyone else’s.
Spend the effort on the fundamentals. Nearly two thirds of the firms in this study were recommended with no file at all, so something else earned them the recommendation. What a firm’s own pages say, whether third-party sources confirm it, and whether the firm’s details are consistent across the web are the working parts of answer engine optimization. CLM applies the same test to every popular claim in law firm marketing, as it did when it measured whether page speed scores affect rankings.
FAQs About llms.txt Helping AI Optimization
What is an llms.txt file?
An llms.txt file is a plain text file placed at the root of a website, for example example.com/llms.txt, that summarizes the site and lists its most important pages for AI tools. It was proposed in September 2024. It is a proposal that websites can adopt voluntarily, and several WordPress SEO plugins can generate one automatically.
Do law firms need an llms.txt file to be recommended by ChatGPT, Claude, Gemini or Grok?
No. In a November 2026 study by Custom Legal Marketing, 63.7% of 3,212 law firm websites recommended by those four platforms had no llms.txt file. Among the 93 firms recommended by all four platforms, 75.3% had none.
Do law firms with an llms.txt file rank higher in AI answers?
The study found no difference in position. Firms with the file averaged position 2.62 in AI answers and firms without it averaged 2.61. In 2,388 answers that included both kinds of firm, firms with the file were listed higher 43.0% of the time and lower 44.8% of the time. Firms with the file were recommended slightly more often on average, 5.04 times against 4.19, a small difference that appeared in only two of six practice groups.
What is the “Hey AI, learn about us” link on law firm websites?
It is a link, usually in the footer, that points AI crawlers to an llms.txt file or to a page written for AI systems. The study found an AI-directed link on 122 of 3,212 recommended law firm websites, or 3.8%. Seventy-five of those firms used some version of the phrase “Hey AI, learn about us,” and 22 used “AI Info” or “AI Information.”
Why does a law firm website have an llms.txt file if nobody at the firm created one?
Many SEO plugins create the file automatically once a setting is turned on. Of the 1,146 llms.txt files found in the study, 574 carried the signature of an SEO plugin. WordPress sites had the file 39.1% of the time, compared with 26.1% of sites on other platforms.
The Fine Print (Disclaimers)
This analysis was conducted by Custom Legal Marketing using its proprietary CLM Sequoia platform. The dataset includes 12,800 prompts in the form “recommend a [practice area] lawyer in [city],” covering 30 practice-area terms and 100 U.S. cities across ChatGPT, Claude, Gemini and Grok, run on November 9, 2026. The prompts returned 55,512 website citations across 7,502 domains. Directories, bar associations, courts, government agencies, legal aid organizations and media were removed, along with nonworking, parked and inaccessible addresses. Websites were visited on November 10, 2026. Site classification was performed using CLM Sequoia with manual review. The analysis set is the 3,212 distinct websites that loaded and were confirmed as working law firm websites.
AI recommendations represent a point-in-time snapshot. Each prompt was run once per platform, and AI platforms can return different firms for the same prompt from one run to the next. Website checks also reflect a single visit, and a file or link present on that day may not have existed when a platform formed its answer. Websites that refused automated visits could not be analyzed and are excluded. Links added to a page only after it loads in a browser may not have been detected, so the count of AI-directed links is a floor. Whether a file was generated by a plugin was determined from signatures inside the file, so that count is also a floor.
The study includes only firms that an AI platform recommended at least once. It does not include firms that were never recommended, so it cannot measure whether an llms.txt file or an AI link affects a firm’s chance of being recommended in the first place. All comparisons describe observed patterns and do not establish causation. Differences between groups were tested with rank-based methods. Across all firms the association between having the file and the number of recommendations was weak (Spearman rho = 0.065, p < 0.001); it was detectable within injury (p < 0.001) and criminal defense (p < 0.001) and was not detectable in family, employment, estate or business groups (p between 0.15 and 0.39). No difference was found in best position (p = 0.87), average position (p = 0.75) or number of platforms (p = 0.61). For the 122 firms with an AI-directed link, the difference in recommendations was not detectable overall (p = 0.41) and was detectable among injury firms (p = 0.03, 63 firms).
© 2026 Custom Legal Marketing, an Adviatech Company. All rights reserved. This research was conducted using proprietary CLM Sequoia technology and methodology.
Jason and his team are the best. He has solid ideas that produce results for the long term. We had tried other companies, but none of them produced the quality results we received from CLM.
- RICHARD WIRTZ
Wirtz Law APC