# Does my business website need an llms.txt file? > llms.txt is cheap to add and there is no public evidence that any major AI provider uses it. What to do with that, without wasting a budget. **Source:** https://parvaly.com/answers/does-my-business-website-need-an-llms-txt-file/ **Author:** Sergey Kalashnikov, Founder, Parvaly **Published:** 2026-08-23 **Topics:** llms.txt, robots.txt, Generative Engine Optimization, GPTBot ## Short answer llms.txt is a proposed standard: a markdown file at your domain root that tells language models what your site contains. The idea is reasonable and the implementation is trivial. What is missing is evidence. No major provider has said it reads the file, crawler logs rarely show requests for it, and nobody has demonstrated a citation outcome from adding one. Publish it if it is free, ignore anyone selling it as the answer, and put the effort into things with measured effects. ## Key takeaways - llms.txt is a community proposal, not an adopted standard, and no major AI provider has confirmed using it. - The file costs essentially nothing to generate, which is the only strong argument for having one. - Server logs from most sites show negligible requests for the file, which is the practical test of whether anyone reads it. - robots.txt is the file that demonstrably matters, because it governs whether AI crawlers may read your pages at all. - Any agency selling llms.txt as an AI visibility deliverable is charging for a hypothesis. ## What llms.txt is supposed to do The proposal is straightforward: publish a markdown file at `/llms.txt` that summarizes your site and points to its most important pages in a clean, machine-friendly form. The reasoning is that HTML is cluttered with navigation and markup, and a model would prefer a curated index. As an idea it is sensible. `robots.txt` and `sitemap.xml` both started as proposals too. ## Why we do not sell it Three things are missing, and they are the three that matter. **No provider has confirmed using it.** Not OpenAI, not Anthropic, not Google, not Perplexity. Adoption of a standard by the parties it is addressed to is not a detail. **The logs are quiet.** The practical test of whether a file is read is whether anything requests it. On most sites, requests for `/llms.txt` are negligible compared with the constant traffic to `robots.txt` and ordinary pages. **No outcome has been demonstrated.** Nobody has published a before-and-after showing citation rates changing because an llms.txt file was added. Until someone does, it is a hypothesis. ## What we do instead, and why The file that unambiguously matters is `robots.txt`, because it decides whether AI crawlers are allowed to read you at all. A detail worth getting right: most providers operate in pairs: a crawler that builds an index, and a separate user-triggered fetcher that retrieves a page live when someone asks a question. Allowing GPTBot but not ChatGPT-User, for example, means your pages can be indexed but not fetched at the moment a customer's question is being answered. Both halves need permitting. Beyond that, the work with observable effects is unglamorous: - Content present in the served HTML rather than assembled by JavaScript, because AI crawlers largely do not execute scripts - A direct, quotable answer in the opening of each page - Strict heading hierarchy, so structure is parseable - Structured data describing the business, its location and its FAQs - A markdown version of each page, if your platform can emit one cheaply. Same idea as llms.txt, but per page and served at the URL a model already has - Consistent business facts everywhere, and third-party sources that corroborate them ## The position, stated plainly We generate llms.txt for the sites we build, because it costs one build step and zero maintenance. We do not list it as a deliverable, do not charge for it, and do not claim it does anything. If someone quotes you for an "AI-ready site" and llms.txt is the headline item, ask them what evidence they have that it works. The answer to that question tells you most of what you need to know about the rest of the proposal. ## Frequently asked questions ### So should I add one or not? Add one if your site generates it automatically at no ongoing cost, which most modern static site setups can do in an afternoon. Do not pay for it as a service, do not expect it to change your visibility, and do not let it displace work that has measurable effects. ### What file actually controls AI crawler access? robots.txt. That is where you allow or block AI crawlers, and it has real consequences. Note that most providers run two agents, one that crawls to build an index and another that fetches a page live when a user asks a question, so allowing only one of the pair leaves you partly invisible. ### Is there anything better to spend that effort on? Yes: making sure your content is in the HTML rather than rendered by JavaScript, putting direct answers at the top of pages, adding structured data, fixing listing inconsistencies, and earning third-party mentions. All of those have observable effects on whether a model can find and verify you. ## Sources 1. [Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-chatgpt-search) · OpenAI --- © 2026 Parvaly LLC. https://parvaly.com