Krolik Marketing

Do I need an llms.txt file to show up in ChatGPT?

No. Google says its search ignores the file, and visit logs from several sites show OpenAI's robots almost never ask for it. Publishing one is harmless. Spend the hour putting your hours, address and services on the page as plain text.

An untouched llms.txt file with cobwebs, while an AI robot reads the plain text on a web page.

What you can do today

  1. Open yourdomain.com/llms.txt in your browser. A website plugin may have made one for you. If it loads, check that its hours, phone and services match your pages.
  2. Check that your site lets OAI-SearchBot in. That is OpenAI's search robot, the program that reads your website for ChatGPT. Here is how. OpenAI has a rule behind that one.
  3. Put your hours, address and services on the page as plain text. This page lists what to include.
  4. In a month, search your visit log for "llms.txt" and see who asked. That is the list your website keeps of every request, robots included.

What the file is

llms.txt is a plain-text list of your pages, written for AI programs. Sellers tell owners they need one.

What Google and OpenAI say

Google's guide, updated 2026-07-10:

"It's completely fine if you decide to create and maintain LLMS.txt files (or other similar files) for other services or systems that use these files. Doing so will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."

That covers Google Search only. It says nothing about ChatGPT.

OpenAI's page about its robots explains robots.txt, the small file that tells robots what they may read. It never says they read your llms.txt. The only mention is a note at the top saying OpenAI's own docs have one.

What visit logs show

Four log studies of llms.txt: 97% of files got no requests, and OpenAI's robots almost never asked, once verified.

These studies come from companies that sell web tools or marketing, so they show patterns, not causes.

  • Ahrefs, 137,210 sites: 97% of the llms.txt files it found got no requests.
  • Arrivl, 39 business sites: going by the name each robot gave, GPTBot (OpenAI's robot that collects text to train its models) made 22 requests, OAI-SearchBot 6, ChatGPT-User 3. Where Arrivl could verify the caller, GPTBot and OAI-SearchBot made none.
  • Rheinwunder, an agency, one server, callers checked by IP address (where the request came from): OAI-SearchBot made 2,432 requests, ChatGPT-User 555, GPTBot 532. None was for llms.txt.

A robot's name can be faked, so counts by name run higher than verified ones. By name, a few requests exist. Verified, almost none.

My own log: 51,427 requests to the site I built for a BMW specialist, Ki-Sol Performance, from 9 October to 15 November 2026.

No named AI robot asked for llms.txt. The only callers were curl, a command-line tool (31 requests), and one unknown robot (2). The file existed when I checked on 16 November. I can't say when it went up.

Arrivl's advice: "Publish it if it costs you an hour, and budget nothing beyond that hour."

If you publish one anyway

Put your name, address, phone, hours and services in it, and keep it identical to your page. That is my advice. No study shows it moves ChatGPT.

I couldn't check whether any OpenAI or Bing product reads the file privately. None of these studies covers auto repair, body shops or churches.

Based on public documentation: Google Search Central; OpenAI documentation; Ahrefs; Arrivl; Rheinwunder.

All posts