
For AI-generated product copy, Merchant Center documents structured text attributes: structured_title for the title and structured_description for the description. Each combines the text with its digital source type. Reviewing the words is still necessary, but approval of the wording does not prove the feed has submitted the required source information.
Key Takeaways
- Keep title and description content connected to their source classification.
- Check the data actually sent by the integration, not only the store editor.
- Image provenance and visible AI labels do not populate text attributes.
Approve the product facts before mapping fields
Google's AI-generated content guide describes an alternative attribute for generated title and description text instead of, or in addition to, the ordinary fields. It specifies the source value trained_algorithmic_media and plain-text content.
First confirm that the wording describes the actual offer. Check identity, materials, dimensions, included items and limitations against the product record. Do not turn a plausible AI sentence into a factual claim merely because it fits a field.
The AI listing-draft review guide covers that content review. Merchant field mapping is the subsequent distribution task, with a different acceptance question.
Keep source and content together
Use the documented structured attribute for each generated text item. The relevant relationship is:
| Text item | Structured attribute | Information to preserve |
|---|---|---|
| Generated title | structured_title | Digital source type and approved title content |
| Generated description | structured_description | Digital source type and approved description content |
The overview is not a universal CSV or API serialization example. Have the integration owner use the current format for the actual data source. Do not paste an invented nested object into a connector that expects a different representation.
For a proposed ceramic-vase listing, the approved description might identify its measured height and actual finish. Keep that exact text associated with its generated source classification, rather than accidentally pairing the field with an older unreviewed draft.
Inspect what leaves the store
A store's title box may display correct words while the export uses only an ordinary attribute. Inspect the relevant outgoing product record or supported integration preview. Record the product identifier, data-source route and both structured components.
If the connector does not expose the required attributes, investigate its supported configuration or contact its maintainer. An AI caption placed on a photograph does not repair a missing text field. Nor does embedded image metadata describe how the separate product description was written.
Keep product promises consistent with the testable-message brief. The image, title and description should not make different claims about the same offer.
Verify the processed product record
After an authorized submission, inspect the affected product's processing result and diagnostics. Compare the sent content with the approved wording, including units and variant details. A successful file transfer is not evidence that every field was interpreted as intended.
Retain the outgoing record with the approved-version handoff. Separate observed field acceptance from any later display or performance result. Close the task when the correct generated copy and source classification reach the intended product, without claiming that structured fields guarantee approval, ranking or sales.
Related tasks
- Is a New Photoroom Product Listing Ready to Go Live After AI Generates It?
- How Do I Turn a Product Launch Brief Into a Testable AI Video Story?
- How Do I Keep the Approved AI Image or Video from Getting Lost Among Revisions?
Sources and scope
- Google Merchant Center: AI-generated content: AI-generated titles use structured_title and descriptions use structured_description, with digital_source_type set to trained_algorithmic_media and the reviewed plain text in content.
Documentation-based guidance from the complete official source captured October 29, 2026. No feed, integration, account or submission was tested. Mapping tables and acceptance steps are proposed; exact serialization depends on the chosen data source. Image metadata is not presented as a substitute for the text attributes.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: AI-generated conceptual editorial illustration; not a real application screenshot, observed output or before-and-after test.