Server logs across major sites show zero requests for llms.txt from GPTBot, ClaudeBot, or Google's AI crawlers. As of late 2025, no major, mainstream AI search engine or consumer chatbot officially supports or relies on llms.txt for crawling, training, or answer generation.
The proposal for llms.txt, introduced by Jeremy Howard in late 2024, gained traction among SEO professionals, developers, and tech-forward sites. But it remains an experimental, non-standardized convention with zero mainstream adoption.
Current status by platform
Google: No AI systems currently use llms.txt. Server logs across thousands of sites show no requests for the file from Google's AI bots.
OpenAI & Anthropic: No evidence that GPTBot or ClaudeBot actively request or use these files for training or retrieval.
Perplexity: Some reports suggest early experimentation, but the file is not a primary (or even secondary) source of data for the engine.
Look the other way
If you're spending time creating llms.txt files, you're optimizing for a future that may never arrive while ignoring the signals that AI systems actually use today.
Creating an llms.txt file is a waste of time. Not because the idea is bad, but because the execution path (adoption by major platforms) doesn't exist.
What AI systems actually use
When Google Gemini cites your brand, when ChatGPT recommends your product, when Perplexity quotes your content, they're pulling from.
1. Entity signals
AI answer engines determine what you are by analyzing consistency across your entire webiste. Not from a text file that says "we are a CRM platform."
2. Content structure
Pages that get cited have clear content, with right HTML formating and little bloat gets chosen. Not because a llms.txt file pointing to a 3,000-word "thought leadership" post with no clear answers.
3. External validation
Brands that get mentioned by name or links acros many different websites and domains gets chosen. Not because they have a llms.txt file saying "we are the leading solution."
What to do instead
Start fixing the signals AI systems actually evaluate.
- Fix entity clarity: Make your brand name, category, and key descriptors identical across every touchpoint: If AI systems see five different descriptions of what you do, they'll cite you without naming you (or skip you entirely).
- Structure pages for extraction: For your 10 most important pages: If a human can't scan your page in 10 seconds and extract the answer, neither can an AI.
- Build external validation: Get mentioned (with context) on sites AI systems already trust: One context-rich mention on a trusted site beats 100 self-referential pages on your own domain.
