Direct answer: For Google Search, an llms.txt file is not a visibility requirement. Google's current generative AI search guide says Google Search ignores these files for ranking and generative AI search: maintaining one neither helps nor hurts Google Search visibility. Another system may choose to use one, so the decision depends on a named consumer and a maintenance plan.
The author's llms.txt proposal describes a Markdown overview with links to deeper, machine-readable pages. It is a proposal for helping agents navigate a site, not a Google Search instruction or a substitute for normal crawlable pages. A file at the root may be simple to generate; keeping it accurate when prices, policies, or product names change is the real work.
Ask three questions before creating the file
1. Which consumer will read it? Name the agent, documentation tool, or workflow and find its current integration documentation or run a controlled test. Publishing a file because “AI likes Markdown” does not identify a reader. Google's guidance provides a clear negative answer for Google Search. Do not infer support by every AI assistant from the fact that some sites publish a file.
2. What is the source of truth? If the overview says a plan includes unlimited seats while the product capability page says ten, the extra file creates a new contradiction. Prefer linking to maintained public pages. If the intended consumer needs a compact index, generate it from the same content inventory or assign an owner to update it alongside the site. Test a few important questions against both the file and the live pages.
3. What work will this replace or delay? Google advises ordinary technical structure and unique, helpful content for its generative Search features, and explicitly calls out unnecessary AI text files as a distraction. If important service pages lack direct links, clear facts, or current terms, repair those first. The sitemap and indexing diagnostic addresses URL discovery; llms.txt does not solve an uncrawlable page for Google.
An optional file is most defensible when a particular agent workflow is documented to consume it and the organization can update it with the website. Record its purpose, owner, generation method, and review trigger. Do not place private support notes or customer data into a public machine-readable file. Review the file after a major product or policy change just as you would the visible site.
This question is separate from crawler permissions for OAI-SearchBot and GPTBot: crawler access determines what a named bot may fetch; llms.txt is a proposed content guide whose use varies by system. An ARCHV website project can implement and maintain such a guide if a real consumer needs it, but it should never be sold as a Google ranking lever.