Traffic analytics cannot capture the full impact of AI-driven search because they only track what users click through, not what AI systems say about a business when no one clicks. This gap is especially visible in local search, where AI tools like ChatGPT and Claude provide direct answers to user queries, often naming multiple businesses without generating any referral traffic.
Most Firms Invisible to AI Answers
In Collin County, Texas, a recent audit of 335 lawyer websites found that only 53 had a dedicated llms.txt file — a plain-text page designed to inform large language models about a business. Of those, 17 were automatically generated by website builders like Wix and GoDaddy, offering little more than generic descriptions such as “trusted legal partner.” Just 36 firms had a meaningful, human-written version.
The same pattern holds for restaurants. Of 503 restaurant sites checked earlier this month, 83 had the page, about 16 percent. Lawyers and restaurants are in the same place.
Wade Kerzie, founder of Kerzie AI Solutions in McKinney, Texas, began investigating after noticing that AI referral traffic accounted for just 0.47% of sessions across client sites. While AI search may not drive significant traffic today, he argues, it already influences which businesses users contact — often without leaving a trace in analytics.
When someone asks ChatGPT for an estate planning attorney who handles blended families, the AI may name three firms. Only one receives the call, and even that visit shows up as direct traffic. The other two firms, plus dozens more not included, remain unaware their names were part of the conversation. There is no public log of which businesses the AI considered and excluded.
Read Also: Canadian fire insurance claims stall over common documentation mistakes
Building a Page That Works for AI
A real llms.txt file includes basic contact details, practice areas described in client-friendly language, and clear information about who the firm serves and who it doesn’t. It lists the counties and courts where it operates, consultation costs, languages spoken, and a note asking the AI to confirm key details with a human before making promises.
Google has stated that such files do not improve search ranking, and Chrome now audits sites for their presence, calling it “a machine-readable summary of a website’s content, specifically designed for LLMs and AI agents.” The purpose isn’t SEO; it’s accuracy in AI-generated answers.
For businesses hoping to show up in those answers, the fix is straightforward: type yourfirm.com/llms.txt into a browser. A 404 means the page doesn’t exist. Typing yourfirm.com/anything.txt and getting the same result as the homepage means the AI is reading the default page instead.
Of the 335 lawyer websites audited, 38 returned a 403 error, blocking AI systems from accessing any content. Ten domains were dead. That leaves a small fraction of businesses with pages that actually help AI systems represent them accurately, and an even smaller number with content that goes beyond boilerplate.
Restaurant Sites Show the Same Gaps
The restaurant audit followed the same method. Of 503 sites checked, 83 returned a valid llms.txt file. The remaining 420 either had no file or served a default page that gave AI systems no useful structure. A handful of chains had files generated by their website provider, listing only hours and contact info.
