Catalogs and menus
Publish and Update a Catalog Without Wasting AI Work
Understand saving, publishing, AI training updates, design-only changes, translation revisions, and unpublishing.
Update the public catalog safely and avoid unnecessary training work.
Saving and publishing are different
Saving keeps the current draft for the owner. Publishing makes that revision available at the public catalog address and connects the current title to the public chatbot navigation.
Use save while editing. Publish when the customer-facing catalog is ready to replace the previous public version.
Only knowledge changes should retrain the AI
Product names, descriptions, categories, tags, prices, visibility, links, and similar customer facts can change what the AI should know. Those changes can queue a training update when published.
Changing only the catalog design should not rebuild vector data. RealLink AI compares the training-relevant catalog content and skips the training job when that content is unchanged.
Translations follow the published revision
A new published revision receives a new content hash. Old translation revisions can be cleaned up, and requested languages are regenerated when the source facts change.
Changing a theme without changing customer text should not create unnecessary translation work.
Unpublish when the page should no longer be public
Unpublishing removes the public catalog connection and removes the system-managed catalog block from AI training. The owner draft remains available for later editing.
After unpublishing, test the old public address and the public chatbot navigation so customers are not sent to a missing catalog.
Quick checklist
- Draft changes were saved before publishing?
- The public page was checked after publishing?
- A design-only change skipped AI retraining?
- Changed facts appear in a new AI test answer?
- The public button disappears after unpublishing?
- Old links were reviewed before turning the catalog off?