The AI analyzes your visual design, detects pages and repeating patterns, and maps them to a CMS structure (collections, fields, bindings).
Where: AI assistant → CMS request — Ask the AI to connect a CMS
Key ideas
- Structure detection: The assistant identifies repeated content regions and proposes collections, typed fields, and repeating/list relationships.
- Visual bindings: Existing text/media elements are mapped to proposed fields so the current design becomes a content template rather than being replaced.
- Reviewable schema: Collection kinds, field names/types, repeaters, relationships, empty states, and remote CMS mapping need human review before publishing.
Steps
- Open a representative content-heavy page and select the repeated list/card region when possible.
- Ask the assistant to propose—rather than immediately publish—the CMS collections, fields, relationships, and bindings.
- Review collection kind, stable field names/types, repeaters, sample entries, and which design element each field controls.
- Test the bound design with empty, single, multiple, long, and missing-value records.
- Connect/publish to a staging CMS only after confirming selective content/layout policies and a rollback path.
Review before accepting
- One collection per durable content type, not per visual card instance.
- Stable machine-friendly field names.
- Correct image/media and relationship types.
- A repeated item container that remains responsive for unknown record counts.
- Explicit empty/missing data behavior.
- No remote write until the proposed mapping has been reviewed.
Limitations and important notes
- Pattern detection can infer the wrong schema from placeholder or inconsistent content.
- Auto-connect does not remove the need to verify remote permissions, field compatibility, and selective publishing behavior.
Troubleshooting
The assistant creates too many collections
Select one representative repeated structure and explain which visual differences are fields/variants of one content type rather than separate types.