Before you begin
Use a workspace with AI access and an existing content model. Free workspaces include 5,000 monthly managed tokens, and Pro includes 100,000 monthly managed tokens plus optional 0%-markup BYOK; your workspace role must also allow the action you are performing. Review your provider settings under Settings → AI if generation is unavailable.
Compare Free, Pro & Max plans or create your first content model.
1. Create a generation run or changeset
Open SCHMA AI in the Studio and start a new generation. Choose the content model, inspect its fields, and enter the instructions for the content you want. On Pro and Max, you can also compile a multi-step changeset spanning interdependent models and sample entries.
Generate a preview to open the staged review workspace. These staged results are not published entries.
2. Review and edit the results
Inspect each staged item, check its field values, and correct the wording or facts. Edits are validated against your model’s field constraints and reference targets. Resolve validation errors before saving an item.
Schema validation checks structure; your editorial review should also cover accuracy, tone, references, and suitability for your audience.
3. Save selected items as drafts
Select the results you want to keep and save them as drafts. Check the outcome for each item: a batch can commit validated items while keeping items needing attention staged so you can edit and retry them.
Saved drafts become available in your content workspace. Review and publish them through the normal publishing workflow when they are ready. Saving AI results never publishes them automatically.
Return to a previous run
Use the searchable run registry to find previous generations, filter by status, or inspect the RFC 8785 canonical changeset hash. Open a run to review its staged results and see what has already been saved. Discard a run when you no longer need its pending results.
When a result cannot be saved
An expired or discarded run cannot be committed. Start a new run if the previous one has expired. If the model has changed, review the current fields and validation messages before saving; if the model was deleted, choose an existing model for a new generation.
For provider or access errors, check AI troubleshooting. Avoid sharing provider keys or database credentials in support messages.