Check for Free Whether ChatGPT Mentions Your Brand

A manual check answers one narrow question: for a handful of buyer questions, today, did ChatGPT name your brand, and was the answer built from live search or from memory?

That is useful. It shows whether you are in the conversation at all, which competitors get named instead, and whether the facts ChatGPT states about you are right.

It does not tell you a rank or a share of voice. ChatGPT has no published ranking you could check a number against, which is why AI rank checkers have no rank to check. Treat the result as a sample, not a score.

This takes about 20 minutes. Use a spreadsheet to record each answer.

1. Remove personal context. Open a private browser window and go to chatgpt.com without signing in. If you must sign in, start a Temporary Chat or turn memory off in Settings under Personalization. OpenAI's help page on ChatGPT search says saved memories can be used when rewriting a search query, so a logged-in test reflects your own history.
2. Write three buyer questions without your brand name. Say you run a payroll tool for restaurants. Hypothetical questions could be "What's the best payroll software for a small restaurant?", "Payroll tools that handle tip pooling" and "Cheaper alternatives to [main competitor] for restaurants".
3. Ask each question in a fresh chat. A new chat stops earlier answers from steering later ones.
4. Record four things per answer: whether your brand is named, which competitors are named, whether sources or citation links appear, and which URLs are cited.
5. Ask one direct question. Try "What is [your brand] and what does it cost?" Compare every fact against your own site.
6. Repeat with two rewordings and on a second day. Rewording shows how sensitive the result is, which matters because phrasing changes whether AI mentions a brand.

For a fuller version across several engines, follow the do-it-yourself AI visibility audit.

Start with the sources column, because it tells you which of two systems produced the answer.

The answer showed sources

ChatGPT searched the web. As of October 2026, it decides per question whether to search, and OpenAI says it rewrites your question into one or more targeted queries before sending them to search providers. If you were named, a fetchable page put you there. If competitors were named and you were not, look at the cited URLs. When engines cite other sites instead of yours, being named on those sites is the route into the answer. That is an association in the evidence, not a guaranteed effect.

The answer showed no sources

ChatGPT answered from training data. That knowledge stops at a cutoff. OpenAI's model documentation lists February 16, 2026 for the GPT-5.6 family and August 31, 2025 for the Instant model behind the chat-latest alias. Anything you published after the cutoff cannot appear in these answers, and nothing you change this quarter reaches them.

The facts were wrong

A wrong price or outdated product name is the most fixable finding. Check that the correct fact is stated plainly on a page that loads without JavaScript. Then see how to tell whether ChatGPT is accurate about your company.

Two people asking the same question can get different answers. That is expected, not a flaw in your test.

• Plan and model routing. Secondary reports say free, Plus and Pro accounts are served different models, and the Instant path has an older cutoff than the reasoning models. OpenAI has not published the mapping itself.
• Search is discretionary. The same question can trigger search once and not the next time.
• Query rewriting. You never see the exact queries ChatGPT sends, so small wording changes can pull different pages.
• Personal context. Memory and custom instructions shape answers for signed-in users.

The practical move: run each question at least twice, log whether sources appeared, and compare patterns rather than single answers.

If ChatGPT shows sources but never names you, check access before anything else. OpenAI's crawler page says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers, though can still appear as navigational links."

Check these in order:

1. robots.txt. Visit yoursite.com/robots.txt. Rules are per bot. GPTBot controls training only, so blocking it does not remove you from search, and allowing it does not help you appear. The search bot needs its own permission:

User-agent: OAI-SearchBot
Allow: /

2. Your CDN or firewall. OpenAI's help page says eligibility also requires your host or CDN to allow traffic from its published searchbot IP addresses. A firewall can return a 403 to a bot that robots.txt allows. Cloudflare changed its AI crawler defaults on September 15, 2026, so review your bot settings even if you edited nothing. Our free AI crawler checker explains how to test this.
3. JavaScript rendering. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but do not execute them. If your pricing or product descriptions only appear after scripts run, the crawler sees an empty page.

The ChatGPT visibility checklist covers each step in more detail. Fixing access makes you eligible. It does not guarantee a mention.

The manual test works, but it is slow to repeat. The free checker on this site runs up to three buyer questions, written by you or drafted from your homepage, and scores the answers. It is a snapshot, not continuous tracking.

For a standalone check, 99Visibility's AI Visibility Checker does the same job. Whichever you use, read results engine by engine. ChatGPT and Gemini draw on different indexes, so repeat the exercise with the free Gemini brand check.

Does allowing GPTBot help ChatGPT mention my brand? No evidence says it does. OpenAI describes GPTBot as the crawler for training its foundation models, and says each crawler setting is independent. OAI-SearchBot controls whether your site can appear in ChatGPT search answers. You can block GPTBot to opt out of training while allowing OAI-SearchBot for search. Whether training inclusion affects what ChatGPT remembers about brands has not been measured by anyone.

How many questions should I test? Three to five buyer questions is enough for a first look, each asked at least twice and with one rewording. More questions give a steadier picture, but no number turns a manual test into a measurement of all real buyer conversations. ChatGPT rewrites questions into its own search queries, and you cannot see those, so any question set is a sample of what buyers might ask.

Why does ChatGPT mention my competitor but not me? In answers with sources, look at which pages were cited. Often they are third-party lists or reviews that name your competitor. Being named on those sites is associated with appearing in the answer. In answers without sources, the model is drawing on training data with a cutoff, and larger, longer-established brands tend to dominate. Neither case has a guaranteed fix, but access problems on your own site are worth ruling out first.

Can I make ChatGPT mention my brand? No one can promise that. What you control is access: OAI-SearchBot permission, a firewall that lets it through, key facts in HTML that loads without JavaScript, and correct facts stated plainly on your pages. Those remove blocks. Which pages ChatGPT selects and cites is decided by its own systems, and OpenAI says only that it ranks results using multiple factors and that placement is not guaranteed.

Is a logged-in test useless? Not useless, but it is personal. Memory and custom instructions can shape how ChatGPT rewrites your search query, so a logged-in result reflects your history as much as the market. Use logged-out or Temporary Chat sessions for the baseline, and treat logged-in results as one extra data point.