Blog
How to track competitor mentions in ChatGPT
Mike Holp · Published · Updated · 4 min read
To track competitor mentions in ChatGPT, run a stable set of unbranded buyer questions, preserve the complete answers and citations, and compare named businesses across repeated samples. This is competitive visibility research, not a hidden ranking report: ChatGPT answers vary by prompt, location, retrieval, model, and time.
Short answer: Choose five to ten buyer questions, keep the location and constraints fixed, run them in fresh conversations, and record whether your business and competitors were mentioned, recommended, or cited. Compare the underlying answers—not just a percentage—because a competitor can be cited as a source without being recommended.
Mention, recommendation, and citation are different
| Observation | What it means |
|---|---|
| Mention | The business name appears somewhere in the answer. |
| Recommendation | The answer presents the business as a suitable choice. |
| Citation | A URL or source is attached to support a claim. |
| Competitor | Another business is named instead or alongside yours. |
| Accuracy | The description matches the real business. |
Treat these as separate fields. A competitor may be mentioned in a warning, cited for a factual detail, or recommended for a specific buyer constraint. Combining them into one “share of voice” score hides the reason an answer changed.
Choose prompts that reveal buying decisions
Avoid prompts that contain your brand name. They measure recognition, not discovery. Write questions a real buyer could ask:
- “What are the best [service] providers for a [customer type] in [location]?”
- “Which [service] is best for a team with [constraint]?”
- “Compare the leading [category] options for [use case].”
- “What should I check before hiring a [provider] in [location]?”
- “Which providers publish clear evidence about [decision factor]?”
Use the same spelling, location, budget, and constraints every time. Change only one variable when you are deliberately testing a hypothesis. Save prompts in a versioned worksheet so an apparent improvement is not just a wording change.
Record a reproducible answer receipt
For every sample, preserve:
- prompt and location;
- engine, mode, and model when exposed;
- run timestamp and sample number;
- full answer text or an approved snapshot;
- every cited URL;
- businesses mentioned and recommended;
- claims that are inaccurate or unsupported;
- provider status, timeout, or partial completion.
OpenAI documents separate crawler purposes for search discovery and model training. If you want a public page to be discoverable in ChatGPT search, inspect the site's policy for the search crawler; do not assume that allowing a training crawler has the same purpose (OpenAI bot documentation).
Compare competitors without over-reading noise
Create a small matrix with one row per prompt and one column per sample. Then ask:
| Question | Your business | Competitor A | Competitor B |
|---|---|---|---|
| Named? | Yes / No | Yes / No | Yes / No |
| Recommended for the stated need? | Yes / No | Yes / No | Yes / No |
| Cited? | URL / None | URL / None | URL / None |
| Description accurate? | Yes / No | Yes / No | Yes / No |
Do not call a competitor “winning” from one response. Mark a pattern only after the same competitor appears across repeated questions or samples, and keep the measurement date. Generative answers are observations, not a permanent SERP position.
Turn a missed mention into a fix list
If your business is absent, first check entity basics: consistent name, canonical website, service description, location, and independent profiles. If it is mentioned but described incorrectly, correct the source page that contains the wrong fact. If it is cited but not recommended, improve the buyer-facing evidence and clarify who the service fits.
Link each fix to a specific observation. “Publish more content” is weaker than “add a page comparing our [service] for [audience], because three samples cited a competitor's comparison page.” The how to get cited by ChatGPT guide covers crawlability and source corroboration.
FAQ
Can I see ChatGPT's competitor ranking?
No. You can observe outputs for a defined prompt set, but there is no universal public ranking dashboard for every ChatGPT answer. Record the exact prompt, context, time, and citations so your report describes an observed sample rather than claiming a hidden rank.
How many prompts should competitor tracking use?
Use at least five distinct buyer questions as a practical starting point, then repeat them across samples. More prompts help cover different intents, but consistency matters more than an arbitrary large number. Label the set and keep it stable between comparisons.
Should I track citations separately from recommendations?
Yes. A citation shows that a source supported part of an answer; a recommendation is a suitability judgment. Track both, along with whether the description was accurate and who appeared instead.
Conclusion
To track competitor mentions in ChatGPT responsibly, use stable unbranded prompts, preserve answer receipts, separate mention from recommendation and citation, and connect each pattern to a concrete content or entity fix. Start with a baseline scan using the VisiScan measurement workflow, then compare like with like.
Sources
- OpenAI: Bots and crawler purposes (reviewed August 2026)
- Google Search: AI features and your website (reviewed August 2026)
Keep going
Turn the ideas in this article into a measurable baseline for your own site.