Quick answer
Separate three cases first: full absence, mentioned but not recommended, or recommended with wrong facts. Remedies differ: strengthen Reach (readable authoritative info), strengthen shortlist/recommendation reasons, or govern facts to fix wrong price/discontinued errors. Do not replace a same-conditions baseline with “ask again and hope.”
A workable path: archive fixed category/comparison prompts on ChatGPT → reconcile public fact consistency → after approval, update fact/product/channel pages → same-conditions retest. Winin’s Edith / Homer / Friday follow that loop.
Details
What not to do immediately
- Treat one screenshot as a formal conclusion
- Bulk-change prices or publish unverifiable “reviews” without approval
- Claim competitors “bought black-hat” without evidence
What to log
Prompt, mentioned/shortlisted/recommended, reason text, cited URLs, competitor set, date. Operating teams can upgrade to a Formal Baseline (scoped quote).
Contact
For help scoping prompts and sampling, email contact@winin.ai.
Related facts
FAQ
Q: Should we manually ask ChatGPT daily?
A: Prefer fixed prompts, archives, and same-conditions retest.
Q: Fix ads or facts first?
A: When facts are wrong, govern facts first; then content and evidence tasks.
Contact: contact@winin.ai · Canonical facts: /en/facts/*/ and https://winin.ai · Updated 2026-09-11