Checking one product? Use the free Product Schema Checker to review brand, GTIN/MPN, price and availability in Product JSON-LD.
Product feed optimization for ChatGPT starts with required fields
For OpenAI’s Stable file-upload discovery feed, the published specification calls for nine basic product fields: item_id, title, description, url, brand, seller_name, image_url, price, and availability. Ads and checkout integrations have separate requirements. A field being present is only the first check; it must also describe the right product or variant.
Where the published feed evidence is accessible to an assessment, Lumen reports observed coverage and highlights which fields need attention first. If a feed value cannot be verified, it should be recorded as unavailable rather than assumed to be correct.
Check each purchasable item and variant
OpenAI’s discovery format expects a row for each purchasable item or variant, with a stable identifier. Review whether the product URL leads to the matching option, and whether the title and description identify it clearly. Keep identifiers as strings so values such as leading zeroes are preserved.
Compare the feed with the live offer
- Identity: the feed row, SKU or variant, and page should refer to the same offer.
- Price: use the current payable amount and the agreed currency; update the feed when a sale begins or ends.
- Availability: keep the feed status current when stock changes.
- Destination and image: use stable, publicly accessible product and image URLs, preferably HTTPS.
- Merchant facts: use the real brand and seller name shown for that offer, not a placeholder.
These checks help teams find the source of a mismatch before changing the feed mapping. A stale inventory export, a generic parent URL, or a variant that inherits the wrong price can undermine otherwise complete product data.
Fix the source, then recheck the same sample
The useful implementation is usually upstream: catalog attributes, variant modeling, storefront templates, structured data, or feed mappings. Record the product sample and observed values first; after the change, repeat the same checks and compare field coverage and page consistency. Lumen’s fix queue is designed to turn verified gaps into implementation work that can be rechecked after deployment.
What a feed audit does not prove
Passing a field-readiness check is not OpenAI approval, ranking, or recommendation. Lumen reports the evidence it can verify and keeps direct-assistant measurement separate from feed-readiness analysis. The assessment does not submit or register a merchant feed on the merchant’s behalf.
Turn feed findings into platform-specific fixes
Once a missing field or feed-to-page conflict is verified, route it to the catalog, theme, structured-data, or feed-mapping layer that owns the value. Use the implementation checklist for Adobe Commerce, Shopify, or WooCommerce. For price, availability, identifier, or selected-variant conflicts, follow the feed-to-page mismatch guide before regenerating the feed.