llms.txt in December 2025: Are Answer Engines Actually Using It?

If you work on SEO, AEO, docs, or AI products, you've probably heard: "You need an llms.txt file or AI will ignore your site."

Short answer as of December 2025:

No major answer engine has publicly committed to using llms.txt in production ranking or retrieval. Some bots poke it. Most ignore it.

This post is a living status page. We'll keep the structure stable so you can skim quickly, and update the details as vendors change their behavior.

1. Quick refresher: what is llms.txt?

llms.txt is a proposed community standard: a Markdown file at /llms.txt (optionally /llms-full.txt) that gives AI systems a clean, structured map of your best content.

Instead of AI having to figure out your whole site from HTML, llms.txt says:

"Here's the important stuff, in Markdown. Here's how it's organized. Here's what's optional."

The proposal is maintained at llmstxt.org.

The llms.txt hub (a community site) describes it plainly as a "proposed standard that helps AI models better understand and interact with your website's content" and explicitly notes that most AI models don't automatically discover or index llms.txt today.

So: it's not an official W3C standard, and it's not a drop‑in replacement for robots.txt. It's a convenience layer for LLMs.

2. About this page (how to read it)

Think of this article as a status dashboard:

• ✅ What the spec says
• ✅ What vendors have said publicly
• ✅ What server logs and audits show "in the wild"

The key summary below is meant to stay fresh; details will shift over time. Everything is "as of December 2025" unless noted.

3. TL;DR - status as of December 2025

Here's the punchline up front:

• llms.txt is still a proposal, not a widely accepted standard. Even Google's John Mueller calls it "just a proposal and not a widely used and accepted standard." (Search Engine Journal)

• Google Search / AI Overviews / Gemini do not use llms.txt. Google reps have repeatedly said:

  • Do normal SEO to appear in AI Overviews.
  • Google will not use llms.txt for AI Overviews ranking. (Search Engine Land) The Search Central docs on AI features talk about structured data, sitemaps, and robots.txt, but never mention llms.txt. (Google for Developers)

• Google's John Mueller: "No AI system currently uses llms.txt. In June 2025 he wrote: "FWIW no AI system currently uses llms.txt" and clarified that none of the major AI services have announced support, comparing it to the old, mostly useless keywords meta tag. (Search Engine Roundtable)

• OpenAI (ChatGPT, GPTBot, OAI-SearchBot) has no official llms.txt support. OpenAI's crawler docs focus on robots.txt and specific user-agents (GPTBot, OAI-SearchBot) with their own allow/deny directives; there's no mention of llms.txt. (OpenAI Platform)

• However, OpenAI does appear to crawl some llms.txt files. Search Engine Roundtable and others have shown server logs where OpenAI repeatedly fetches /llms.txt for some sites - roughly every 15 minutes - suggesting experiments, not an announced product feature. (Search Engine Roundtable)

• Anthropic and Google both publish llms.txt for their own docs.

• Anthropic's engineering blog explicitly recommends using flat llms.txt documentation files (and links to its own) so Claude-based tools can load docs efficiently. (Anthropic)

• Google's Gemini API docs are exposed via an llms.txt endpoint, similar to how many dev-doc sites now do it. (Google AI for Developers)

• Perplexity and other AI search tools also publish llms.txt for their docs. The llms.txt hub lists Perplexity's llms.txt/llms-full.txt alongside Anthropic, Cloudflare, ElevenLabs, etc., as implementers of the proposed standard. (llms.txt hub)

• But none of these vendors say their answer engines consume arbitrary sites' llms.txt in ranking or retrieval. Multiple reviews (Ahrefs, Publii, others) point out that no major LLM vendor has publicly committed to honoring llms.txt in their crawlers or training pipelines, and there's no evidence it improves visibility in AI answers yet. (Ahrefs)

• Large log studies and experiments mostly show AI bots ignoring it.

• A 1,000‑domain CDN log audit found no hits from GPTBot, ClaudeBot, or PerplexityBot to /llms.txt; the vast majority of requests came from Googlebot, Bing, and SEO tools. (Flavio Longato)

• A live experiment tracking hits from GPTBot, ClaudeBot, and PerplexityBot also reports very sparse traffic so far. (Complete SEO)

• Adoption by websites is growing anyway. Tools like BuiltWith show 2,000+ websites exposing an llms.txt file (many dev-doc and SaaS sites), up from almost none in early 2024. (getairefs.com)

Net-net:

As of December 2025, most major answer engines do not use llms.txt in any guaranteed way, but some bots are starting to crawl it, and many documentation-heavy sites have adopted it "just in case."

4. Real‑world evidence: what server logs show

Beyond vendor statements, we have several independent looks at how bots behave:

• A 30‑day log audit of 1,000 enterprise domains found:

  • No requests to /llms.txt from GPTBot, ClaudeBot, PerplexityBot or similar.
  • ~95% of llms.txt hits came from Googlebot desktop, with a handful from Bing and various SEO tools. (Flavio Longato)

• The Publii "Complete Guide to llms.txt" summed up the landscape (Oct 2025) as:

  • Zero official platform adoption
  • No confirmed cases of LLMs using llms.txt to improve responses
  • No peer‑reviewed evidence of impact (StaticCMS)

• Airefs' explainer notes:

  • >2,000 sites already expose llms.txt.
  • There are "early reports" of GPTBot crawling the file, but nothing that proves large‑scale consumption. (getairefs.com)

• A live experiment by Complete SEO tracks hits to llms.txt from GPTBot, ClaudeBot, and PerplexityBot. So far, traffic appears sporadic and thin, reinforcing the idea that this is still early and experimental. (Complete SEO)

So the pattern is:

llms.txt is being published a lot. Traditional web crawlers (Google, SEO tools) are reading it a lot. AI‑specific bots are mostly ignoring it, with a few signs of emerging tests.

5. Should you ship llms.txt in 2025?

Here's the practical bit:

llms.txt is a cheap future‑proofing bet in case vendors start honoring it.

A good mental model for December 2025:

It won't do anything now, but it's a one-time, low-effort, safe change. If it's easy for you - do it, but keep your expectations very low.

If you do implement it, keep it simple. Follow the spec from llmstxt.org, and consider the hub's advice:

• H1 title, short summary.
• A few sections (Docs, API, Tutorials, Examples).
• URLs pointing to Markdown or clean HTML where possible. (llms-txt)

Monitor logs, not just hype. Check whether GPTBot, ClaudeBot, PerplexityBot, or others are actually touching /llms.txt on your domain. Use that to guide how much time you invest going forward. (Flavio Longato)

Final takeaway

As of December 2025:

llms.txt is real, growing, and sometimes crawled - but still not a dependable way to influence how major answer engines like Google AI Overviews, ChatGPT, Claude, Perplexity, or Gemini see your site.

Worth implementing if it's easy. Not worth losing sleep over.