We Tracked 10 Sites: Does llms.txt Matter?
▼ Summary
– The debate around llms.txt is highly polarized, with some treating it as essential infrastructure while many SEO experts dismiss it as speculative, and server logs show AI crawlers rarely request these files.
– A study tracking 10 sites found that implementing llms.txt did not cause AI traffic growth; the two sites that saw increases had launched new, functional content and made other significant improvements concurrently.
– No major AI service provider, including OpenAI, Anthropic, Google, or Meta, has officially committed to using llms.txt, and Google’s John Mueller stated they do not check for it.
– The primary practical value of llms.txt is token efficiency for parsing, which is relevant almost exclusively for developer tools and API documentation where AI coding assistants are a key channel.
– For most websites, llms.txt should be treated like a sitemap, useful infrastructure but not a growth strategy, with effort better spent on creating extractable content, fixing technical issues, and earning external validation.
The discussion surrounding llms.txt files has grown increasingly divided within the digital marketing community. Some view it as essential new infrastructure, while experienced SEO professionals often regard it as an unproven experiment. While platform tools may flag a missing llms.txt as a site issue, server logs consistently show that AI crawlers from major providers seldom request the file. Even Google’s brief, partial adoption of the standard on some properties was quickly reversed, with a company representative clarifying the files served other internal purposes and were not intended for AI discovery. This leaves webmasters questioning whether implementing llms.txt is a strategic necessity or a distraction.
To move beyond speculation, we conducted a 90-day tracking study across ten websites in diverse sectors including finance, SaaS, ecommerce, and insurance. We monitored AI crawl behavior and traffic from platforms like ChatGPT and Gemini both before and after llms.txt implementation, while also noting other concurrent site changes. The findings were revealing: eight of the ten sites experienced no measurable change in AI traffic. One site saw a decline of 19.7%, a drop that correlated with broader site trends unrelated to the text file. Only two sites posted gains of 12.5% and 25%, but deeper analysis showed llms.txt was not the catalyst.
In the case of the digital bank that grew 25%, the period also featured a major PR campaign for a new banking license, a rebuilt resource center, the publication of twelve new FAQ pages, and fixes to critical technical SEO issues. Similarly, the B2B SaaS company’s 12.5% spike coincided with the launch of 27 functional, downloadable AI templates, a new site section that also saw an 18% rise in Google organic traffic. The clear pattern was that sites launching new, functional, and well-structured content saw gains, while sites that merely documented existing content in an llms.txt file saw no impact.
The core issue is a lack of official support. No leading large language model provider, OpenAI, Anthropic, Google, or Meta, has committed to using llms.txt for discovery. As Google’s John Mueller noted, you can verify this by checking your server logs; the crawlers simply aren’t looking for it. The most compelling argument for the file is token efficiency, particularly for developer tools and API documentation where clean markdown can streamline how AI coding assistants parse information. For most other businesses, however, this efficiency doesn’t translate into tangible traffic benefits.
Ultimately, llms.txt functions more like a sitemap than a strategy. A sitemap is valuable infrastructure that aids discovery, but no one attributes traffic growth solely to its existence. It documents content but doesn’t enhance its quality or usefulness. Our research underscores that what truly works for AI discovery mirrors proven SEO fundamentals: creating functional, extractable assets like templates and comparison tables, structuring content clearly for machine parsing, eliminating technical crawl barriers, and earning external authority through press and backlinks. These are the levers that drove results in our study.
So, should you implement an llms.txt file? For developer tools where AI agents are a key distribution channel, it may be a prudent step for future-proofing and efficiency. For everyone else, it’s best treated as optional infrastructure, it likely won’t hurt, but it almost certainly won’t drive growth. The time invested is typically better spent on activities with a proven return: optimizing product pages with structured data, publishing actionable content, resolving technical issues, or building authority. The lesson isn’t that llms.txt is without potential, but that in a landscape with unwritten rules, focusing on creating useful, accessible, and well-structured content remains the most reliable path to visibility, regardless of how platforms evolve.
(Source: Search Engine Land)
