How to Tell Whether ChatGPT Actually Uses Your Website

Whether ChatGPT uses your website comes down to three separate things: can it reach your pages, does it cite them when it answers, and does it name your business? A site can pass the first and fail the other two. The only reliable way to know is to check each one directly: your server logs for the first, and the answers themselves for the rest.

Readable, cited and named mean different things

It's easy to treat "does ChatGPT use my site?" as one question, but it's really three. Readable means its crawlers can fetch your pages. Cited means it lists your page as a source in an answer. Named means it mentions your business in the answer itself. A page can be cited without your business being named, and a business can be named from a directory listing without its own site being cited at all.

If you haven't checked the first one yet, start with whether ChatGPT can read your website.

Check 1: Is it reaching your pages?

OpenAI uses identifiable crawlers (GPTBot for training and OAI-SearchBot for search) and a user-triggered agent, ChatGPT-User, when someone asks it to open a page. Your server or hosting access logs record these by user agent. If they never appear, or appear with error responses, ChatGPT likely isn't getting your content. A blocked request usually points to robots.txt or a firewall rule. How to spot a firewall block.

Check 2: Does it cite your site?

When ChatGPT searches the web to answer, it can show the sources it drew on. Ask it the questions your customers ask, with web search in play, and look at the sources. If your pages appear, your content is being used. If a competitor's or a directory's pages appear for questions about your own services, that's the gap to close.

Why the sources you see can vary

ChatGPT doesn't search the web for every question, and when it does, the pages it reads can differ from one run to the next. So your site may be cited in one answer and missing from the next, even with nothing changed on your end. That's normal, and it's another reason to judge citation by how often it happens across repeated checks rather than by any single answer.

Check 3: Does it name your business?

This is the outcome that leads to phone calls. Look at the answer text as well as the sources: is your business mentioned by name, and is it recommended or just listed? A cited page that never gets your name into the answer usually means your identity isn't coming through clearly. It's worth tightening your structured data and making sure your business name appears plainly on the page.

What to do with what you find

Each check points to a different fix. If OpenAI's crawlers aren't reaching you, look at robots.txt and firewall settings first, because the other fixes won't help until they can get in. If you're reachable but never cited, your pages may not answer the questions directly enough, or others answer them more clearly. If you're cited but not named, strengthen your identity on the page: your business name, location and services stated plainly, with LocalBusiness structured data behind them.

A note on JavaScript-heavy sites

Many modern websites load their content with JavaScript after the page first arrives. Not every crawler runs that JavaScript, so a site that looks complete in your browser can look almost empty to some of them. If your site is built this way, check that your key text (services, locations, contact details) is present in the page's initial HTML. More on readability.

Why you have to check more than once

ChatGPT's answers vary from one run to the next, even for the same question. One check can catch you on a good day or a bad one. Ask the same questions repeatedly over a few weeks and look at the rate. Speak Local does this across ChatGPT, Perplexity and Gemini, and records every answer as named, cited, read or absent, so the pattern is visible.

How to run the checks yourself

  1. Ask your hosting provider or developer how to read your access logs, then search them for "GPTBot," "OAI-SearchBot" and "ChatGPT-User." Note the response codes they received.
  2. List the five questions customers most often ask before calling you.
  3. Ask ChatGPT each question, phrased naturally and including your town, and note the sources shown and the businesses named.
  4. Repeat the questions a week later and compare.

That's enough to tell you which of the three checks you're failing. Doing it by hand every week across several assistants gets tedious. How to keep a ChatGPT check going week to week.

Quick answers

How can I tell if ChatGPT has crawled my site?
Look in your server or hosting access logs for OpenAI's user agents: GPTBot, OAI-SearchBot and ChatGPT-User. If they're absent or receiving error responses, something is likely blocking them.
Why is my site cited but my business not named?
The assistant used your page as a source but didn't carry your business name into its answer. Making your name, location and services explicit on the page, and adding LocalBusiness structured data, helps your identity come through.
Can I ask ChatGPT whether it uses my site?
You can ask, but its answer about itself isn't reliable evidence. Checking your logs and looking at the sources and names in real answers to real questions is far more trustworthy.

Scan your own site

Everything on this blog is measured by the Speak Local scan. Run it on your own domain and it reads your pages the way Googlebot, GPTBot, ClaudeBot and PerplexityBot do, then scores what it finds.

Free, in under a minute. No account.  Or start a plan from $49/mo

Google SearchChatGPTPerplexityGeminiAI Overviews