3. How do I check whether ChatGPT recommends my company?

Run a small, disclosed set of unbranded buying questions in fresh conversations and retain the complete answers. Keep ordinary-use observations separate from tests where you explicitly require web search.

OpenAI documents that ChatGPT can rewrite requests into search queries and that saved memories and location information can influence search. A fresh conversation therefore does not, by itself, establish an unpersonalised or location-neutral test. (help.openai.com)

For a first diagnostic, we propose:

1. Select six buying questions covering discovery, requirements and shortlisting.
2. Specify the intended US customer, language and geographic scope.
3. Test two relevant surfaces, one of which may be ChatGPT.
4. Collect three scheduled observations per question per surface: 36 planned observations.
5. Record mentions, owned-site citations and explicit recommendations separately.
6. Save timestamps, full answers, source URLs and collection failures.
7. Have a person review ambiguous recommendations and company-name matches.

Record the product surface, account tier, visible model label where available, search setting, memory settings and location configuration. Do not invent a model version when the interface does not disclose one.

This is a triage exercise, not a statistically validated market benchmark. Use it to decide which questions deserve deeper measurement—not to claim that a particular percentage of US buyers will encounter your brand.