OpenAI warns buyers to verify
OpenAI’s Shopping Research announcement notes that price and availability can still be wrong and should be confirmed on the merchant site. That limitation is important because commerce facts change faster than descriptive content.
Source: ChatGPT Shopping Research.
There are several clocks
Price or stock can exist in:
- Shopify variant record
- public page HTML
- JavaScript-selected variant
- Product
Offerin JSON-LD - product feed
- Shopify Catalog
- search or retrieval cache
- generated answer
These sources do not necessarily update at the same second.
Diagnose the exact variant
If ChatGPT reports USD 89 for a product that now costs CAD 129, capture:
- prompt and timestamp
- cited URL
- country and language
- product and selected variant
- displayed price and currency
- page canonical
- current feed record
Do not investigate “the product” in general. A size, color, subscription option, or market may carry the stale value.
Common Shopify causes
| Symptom | Likely check |
|---|---|
| sale price persists | compare-at and feed refresh |
| sold-out item appears in stock | duplicate or stale Offer |
| wrong currency | market redirect and destination URL |
| subscription price shown as ordinary | selling-plan labels and feed mapping |
| one size price used for all | variant selection and item grouping |
Reduce the error window
Use stable variant IDs. Keep inventory updates flowing to connected channels. Remove duplicate Product objects. Ensure the canonical destination presents the same market and purchasable state as the catalog record.
When a product changes materially, record the time it changed and the time each connected channel accepted the update. This is data freshness, not merely content quality.
Report accurately
Say:
At 14:20 MT, the cited page showed CAD 129 and the answer stated
USD 89. The source and market did not agree.
Do not say:
ChatGPT always has the wrong price.
For the technical repair, use Shopify price differs between page and JSON-LD.
Community discussion
Add to the article
Ask a technical question, share a storefront result, or challenge a conclusion with evidence.