Maryland is changing how residents and AI tools find government information by publishing a single plain-text file that tells chatbots how to crawl and use official content.
The file, known as llms.txt, is intended to help automated systems locate the most reliable sources on state websites and avoid pulling from temporary or misleading pages. State officials say the move is low-cost but can reduce errors when virtual assistants respond to public queries.
What this means for residents
For Marylanders, the change should make chatbot answers about state services, forms and guidance more accurate. The llms.txt file directs AI agents away from transitory items such as emergency banners and short-lived alerts that could otherwise produce outdated responses.
David Holmes, Maryland’s senior director of engineering, described the file as a way to improve people’s digital experience. He told Technical.ly the file helps ensure AI connects users with relevant information at the time they ask.
“We’re just doing our part to make sure AI can get the most relevant information at the time.”
Scale and scope
State staff report the Maryland website receives tens of millions of AI-related requests monthly and a significant portion come from chatbot-style agents. Key figures cited by state officials include:
- 40 million to 50 million AI crawler requests per month across the state web portfolio.
- About 6 million of those tied specifically to chatbot queries.
- 21,400 requests for llms.txt files recorded over a recent 90-day period.
| Metric | Reported value |
|---|---|
| Monthly AI crawler requests | 40–50 million |
| Monthly chatbot queries | ~6 million |
| llms.txt requests (90 days) | 21,400 |
How llms.txt works
The plain-text file maps site sections and flags content types that should not be summarized or used as authoritative answers. Among the items Maryland’s file discourages AI from using are emergency alerts, temporary banners and pages that invite legal interpretation.
Holmes said the approach functions like "air traffic control for bots," steering automated agents to the most relevant pages. He framed the effort as a straightforward improvement to the public’s digital access to government services.
“We’re thinking, ‘How do we make the digital experience for folks better, no matter where they’re at?’”
Context and limitations
Among web developers and AI experts, opinion is mixed about llms.txt’s long-term usefulness. Major AI firms have not formally committed to honoring the format, and many crawlers still rely on standard HTML and existing protocols like robots.txt. Nevertheless, Maryland officials view the file as a practical step to reduce misinformation and improve chatbot behavior without heavy technical investment.
As residents increasingly use assistants such as ChatGPT and Claude to find government information, state technology leaders say simple guardrails like llms.txt can help ensure those tools connect users with accurate, up-to-date sources.
Maryland’s move makes it the first state known to publish such a file for its official website, a distinction officials hope will serve as a model for other governments weighing how to regulate AI access to public information.
