For a standard website, the llms.txt file provides no measurable benefit in 2026, neither in Google nor in AI search engines, according to Abondance. The site analyzed its server logs between June 1 and November 4, 2026: its two files received 1,657 requests, of which 10 came from identified AI bots.
What Google and Studies Say
The llms.txt is a Markdown file (plain text format) placed at the root of a website. It introduces the site and lists its useful pages. This format was conceived by Jeremy Howard and presented on September 3, 2024. Two years later, it is still not a web standard. A variant, llms-full.txt, groups the complete content of pages into a single file. It is not part of the original format and comes from documentation tools like Mintlify.
Google states, in its guide on generative AI, that no text file is necessary to appear in search, including AI features. The search engine specifies that “Google Search ignores them.” In a podcast dated June 15, 2026, John Mueller believes that hoping to be discovered by AI thanks to this file “doesn’t make any sense at all,” as every site presents itself as the best.
Studies point in the same direction. Ahrefs analyzed 137,210 domains in May 2026: 28% publish an llms.txt, and 97% of these files received no requests in the month. SE Ranking, for its part, cross-referenced nearly 300,000 domains with their citations in LLMs and found no correlation with the presence of the file.
What Abondance’s Logs Show
Of the 1,657 requests recorded, 585 come from SEO, GEO, and scanner tools, accounting for 35% of the total. An OVH server impersonating Googlebot opened the file 289 times since October 23, while the real Googlebot only visited 34 times during the period. ChatGPT-User, which only operates upon user request, opened the file twice, on June 8 and August 25. ClaudeBot opened it once, on October 21, and OAI-SearchBot, PerplexityBot, and GPTBot never consulted it.
An update published on November 5 adds an important clarification. Abondance’s host discovered that GPTBot had been blocked since August 25 by its firewall, which contained addresses of the bot and some addresses of PerplexityBot. The list was removed the same day, and both bots reappeared in the logs. Therefore, allowing a bot in robots.txt does not guarantee it will reach the site, as a firewall or CDN can block it upstream.
Cases Where the File Remains Useful
John Mueller highlights technical sites in particular. Google itself publishes this type of file on its developer sites. In the Ahrefs study, Claude Code, Anthropic’s coding agent, is among the AI tools that open llms.txt files the most. For a service that an agent needs to use, the file can explain how to proceed, although John Mueller notes that the question remains open.
