In a move that has baffled the SEO community, Google has uploaded an LLMs.txt file to its Search Central portal—months after claiming the file was useless for indexing. While Search Advocates advised webmasters to noindex these files, Google is now using one to help AI models parse its own documentation. Is this a case of “do as I say, not as I do,” or a sign of an upcoming AI search shift?
Here is the breakdown of the contradiction you pointed out, detailing exactly what Google said versus what they just did.
1. What Google Said (“Don’t bother”)
For months, Google’s search relations team has actively discouraged site owners from spending time on llms.txt files.
- “It’s Useless”: John Mueller (Google Search Advocate) and Gary Illyes (Google Search Central) have publicly stated that Google Search does not use the file for indexing or ranking.
- “No One Uses It”: They claimed that currently, no major AI systems or LLMs (including Gemini) actually respect or look for the file in a meaningful way for discovery.
- “Noindex It”: Their specific advice was that if you do create one, you should slap a
noindextag on it. Why? Otherwise, Google Search might index the text file itself, and it would appear in search results as a broken or low-quality page, cluttering the index.
2. What Google Did (“We’ll use it”)
Despite all the above, users recently discovered that Google has uploaded its own llms.txt file for the Google Search Central documentation.
- The File: You can view it at
https://developers.google.com/search/docs/llms.txt. - The Content: The file follows the standard proposal format. It contains a clean, Markdown-based list of their documentation, explicitly structured to help LLMs digest their technical guides more easily.
- The Irony: They are using the exact file format they told webmasters was “useless” to ensure that LLMs (like ChatGPT, Claude, and Gemini) can better understand their own documentation on ranking on Google.
Why the Double Standard?
There is no official explanation yet, but the community has a few theories:
- Internal Testing: The team managing
developers.google.comMight be testing the standard independently of the Search team. Google is a massive company, and the “Search” team (which talks to SEOs) is different from the “Docs” team. - Future-Proofing: They may know that support for
llms.txtGemini or Search soon, and they are quietly preparing their own assets first. - “Trolling”: Some in the SEO community have half-jokingly suggested this is a troll move, given the timing of their public dismissal of the file.
The Takeaway
For now, the official advice for you remains the same: Google Search does not support it.
However, Google’s actions suggest they see value in it for other LLMs (such as OpenAI’s crawlers) to understand their content better. If you have a content-heavy site (like documentation or a blog) and want to make it easier for non-Google AI agents to read your site, following Google’s actions rather than their words might actually be the right move.
While Google maintains that LLMs.txt remains unused for traditional search ranking purposes, their implementation of the file on the Google Search Central portal sends a mixed signal that webmasters cannot ignore. Whether this is a case of internal miscommunication between the Search relations team and the documentation engineers, or a quiet pilot program for future AI discovery, the takeaway is clear: the landscape of AI crawlers is evolving. SEO professionals should view this as a potential pivot point. Even if Google Search doesn’t currently honor the file, other AI agents and LLMs—such as those powering ChatGPT and Claude—rely on these files for context. Consequently, adopting this web standard acts as a crucial step in future-proofing your content for the broader generative AI ecosystem. As the line between traditional search and LLM-based answers blurs, following Google’s actions rather than their words may be the smartest play for maximum content visibility