To track ChatGPT brand mentions reliably, build a fixed set of 20 to 30 commercial buyer prompts, test them biweekly in clean unpersonalized US sessions (or via an automated tracker), and log four fields separately: brand mention presence, shortlist position, linked domain citations, and pricing or feature accuracy.
1. Separate Brand Mentions From Source Citations
ChatGPT does not report visibility in one uniform format. Sometimes the model names your company inside the recommendation list without linking to your website. Other times it cites one of your guides in a footnote without naming your product in the main answer.
Tracking both numbers separately prevents false conclusions. High mentions with low citations usually mean ChatGPT learns about your brand from third-party reviews rather than your own documentation.
2. How to Track ChatGPT Mentions Manually With 20 to 30 Prompts
Before buying dedicated software, a structured spreadsheet lets you inspect exact wording, competitor framing, and cited URLs firsthand.
- Collect 20 to 30 commercial prompts from sales calls, support tickets, and category comparisons (see our AEO prompt selection framework).
- Disable chat history or use a fresh, unpersonalized US browser session with web search enabled so past chats do not bias recommendations.
- Log six columns per prompt: Date, Prompt, Brand Mentioned (Yes/No and rank #1-#5), Owned URL Cited, Sentiment/Accuracy, and Competitors Named.
- Repeat the exact same prompt set every 2 weeks rather than judging performance from a single snapshot.
3. When to Switch From Spreadsheets to Automated Tracking
Manual logging works well for one product line and 20 prompts, yet it breaks down quickly once you track multiple buyer personas, US regions, or weekly competitor shifts. Automated trackers run the same prompt set on a schedule, extract cited domains, and calculate AI share of voice over time.
To run an immediate baseline without setup overhead, test any domain on 5 live non-branded US buyer prompts inside our ChatGPT Visibility Tracker.
4. Turning ChatGPT Tracking Data Into Fixes
Each combination of mentions and citations points to a different root cause and next step:
| Tracking Outcome | Likely Root Cause | Next Action |
|---|---|---|
| Mentioned + Cited | Strong owned specs and off-site consensus | Keep pricing and modified dates current |
| Mentioned, Not Cited | ChatGPT relies on third-party sites for facts | Publish plain-HTML comparison and spec tables |
| Cited, Not Mentioned | Educational content ranks, product fit unclear | Clarify target buyer, pricing, and use cases |
| Neither Mentioned Nor Cited | Blocked crawler or weak category presence | Verify OAI-SearchBot in robots.txt and G2 profiles |
Frequently asked questions
Why does ChatGPT give different brand lists when I run the same prompt twice?
LLMs sample responses probabilistically and fetch live web results that update continuously. Testing a fixed 20-to-30-prompt set biweekly smooths out single-run variance.
Does ChatGPT Search use my chat history when recommending brands?
Logged-in sessions with memory or custom instructions can bias recommendations toward brands you previously discussed. Always run visibility checks in clean, logged-out or unpersonalized sessions.
Can I see ChatGPT referral clicks in Google Analytics?
Yes, for users who click footnote links. Configure a custom AI channel group in GA4 as shown in our GA4 AI referral tracking guide.
Next steps & related pages
Want a baseline visibility check for your brand?
Save the CSV checklist above for manual audits, or request a baseline check on your target US buyer prompt while our automated tracking workspace rolls out.