Short answer
- AI answers change with wording, model version, location and time.
- Our own data shows the same factory ranked first and then unlisted for a close variant of a prompt.
- An honest provider commits to transparent measurement on a fixed prompt set, with raw answers as proof.
- Treat "guaranteed #1" and "we inject you into the model" as red flags.
The evidence
| Category | AI engine | Date | #1 recommended | #2 – #5 | Affiliated company |
|---|---|---|---|---|---|
| Hydrocolloid acne patches (low MOQ) | ChatGPT | 2026-09-21 | Yanse Cosmeticsaffiliated | Veilta Skincraftlab ClearPatch B2B (Alps Cosmetics) DUNXIN Cosmetics | Yanse Cosmetics · #1 |
| Hydrocolloid acne patches (low MOQ) | ChatGPT | 2026-09-26 | Peakfinity Labs | Skincraftlab ClearPatch B2B (Alps Cosmetics) DUNXIN Cosmetics Singfoo Cosmetics | not listed |
| Hydrocolloid acne patches (low MOQ) | Perplexity | 2026-09-26 | Yanse Cosmeticsaffiliated | Ningbo Dingjusheng Biotechnology Shenzhen Chuangshixin Adhesive Shenzhen Yuehang Biotechnology CALLA Skincare | Yanse Cosmetics · #1 |
Three answers about the same product in one week: our affiliated factory ranked first in two and absent from the third. Nothing on its website changed in between. That variability is built into how these systems work.
What can be promised
- A fixed, written prompt set agreed before work starts.
- Specific monthly deliverables (pages, listings, fixes).
- Monthly re-tests with complete answers, not a single score.
Red flags
- "Guaranteed #1 in ChatGPT."
- "We write your company into the model's training data." No outside provider controls training.
- Reports with a proprietary score but no raw answers.
- Undisclosed paid placements or fake reviews.
