Wrote an llms.txt validator with 64 tests — and checked whether bots read this file
The author created a generator and validator for llms.txt, ran 64 tests, and found that AI bots rarely request the file: only 408 requests out of 515 million bot events. The validator helps ensure the file meets the specification, which is often violated.
An article on Habr describes the development of a validator and generator for the llms.txt file, a proposed standard for providing site summaries to language models. The author conducted 64 tests and discovered that AI bots rarely request llms.txt: Limy.ai's May 2026 analysis showed only 408 requests out of approximately 515 million bot events over 90 days. Despite this, the validator proved useful because the specification is often violated in real files, with issues like missing titles, multiple H1s, and http links. The parser separates parsing from validation, and the validation score subtracts weighted penalties (critical 40, high 25, medium 12, low 5) from 100. The test suite covers 22 specific rules and edge cases, running in 43 milliseconds. The author recommends using llms.txt as a cheap insurance policy rather than a guaranteed path to AI answers, and advises checking the file structure before publishing.
- Abbreviations
- llms.txt = Large Language Models text file — текстовый файл для языковых моделей
- H1 = Heading level 1 — заголовок первого уровня
Source: Habr — хаб ИИ —
original