Google\u2019s John Mueller says markdown pages serve a specific purpose for developer documentation sites but won\u2019t help most websites, even as search becomes more agentic.<\/p>\n
Mueller laid out his reasoning in a Bluesky thread<\/a>. He was responding to a question from Lily Ray about why Google publishes LLMs.txt files and markdown pages, even though they aren\u2019t needed for search performance.<\/p>\n His response focused mainly on markdown versions of developer documentation, not llms.txt as a standalone file.<\/p>\n Mueller wrote:<\/p>\n \u201cThe short answer is that it\u2019s not done for search. There\u2019s more to websites than just SEO :-).\u201d<\/p>\n Mueller\u2019s Discovery Vs. Functionality Framework<\/p>\n His reasoning focuses on two different website goals.<\/p>\n He called the first \u201cdiscovery,\u201d or being found via a search engine, and the second \u201cfunctionality,\u201d which helps users complete tasks on the page.<\/p>\n Mueller acknowledged the term wasn\u2019t precise. \u201cThere\u2019s probably a more accurate term for this,\u201d he wrote in the thread.<\/p>\n He compared the distinction to calls to action on traditional pages, stating:<\/p>\n \u201cYou don\u2019t \u2018do them\u2019 for SEO (to be found), but if you\u2019re responsible for the website overall, ensuring a high \u2018discovery rate\u2019 (SEO) together with a high conversion rate is useful to justify your work.\u201d<\/p>\n Why Developer Docs Are Different<\/p>\n On developers.google.com, he noted, markdown versions make sense.<\/p>\n Mueller said;<\/p>\n \u201cAI coding has gotten very popular, and these coding systems can be (I think) efficient and accurate with the code they produce if they can easily read \/ parse reference material, such as developer documentation.\u201d<\/p>\n He added that markdown can help AI systems \u201cunderstand the context of the documentation they\u2019re looking at, as well as a simplified version of the reference page.\u201d<\/p>\n Mueller called this a workaround rather than a long-term need, adding:<\/p>\n \u201cOF COURSE they can read HTML just fine, so this is imo more of a temporary crutch, perhaps to save some tokens.\u201d<\/p>\n Non-Developer Sites Should Skip It<\/p>\n For everyone else, Mueller was direct, stating:<\/p>\n \u201cFor non-developer sites, I don\u2019t think this makes much sense, even with more agentic traffic in the future. Making a markdown version of a shoe\u2019s specs is not going to get you more sales (competitors appreciate it tho).\u201d<\/p>\n He went further in a follow-up post, pushing back on the idea that sites should prepare for a future where agents drive more traffic.<\/p>\n Mueller added:<\/p>\n \u201cAnd (I know, nobody reads this far), if you think this is important to prepare for when agents are everywhere: your site (all sites) have much more important things to do for SEO than to prepare for a potential future situation that may or may not come. Prioritize needs before dreams.\u201d<\/p>\n Why This Matters<\/p>\n Mueller\u2019s comments show a more detailed position than his earlier statements on the topic.<\/p>\n In February, Mueller called the idea of serving markdown pages to bots<\/a> \u201ca stupid idea.\u201d His Bluesky comments carve out an exception for developer documentation while holding the line for every other type of site.<\/p>\n