What it means
A schema is the contract between prompt, model output, and interface: which fields exist, what they mean, and what “valid” looks like.
Why designers should care
Without a shared schema, teams get brittle prompts and broken UI when the model omits fields, renames keys, or drifts format across releases.
Example
A user-research synthesizer schema requires theme, evidence quote, severity, and recommendation. Missing fields show inline “Regenerate this section” instead of silent gaps.
Common mistakes
- Letting the model invent field names per request with no stable schema.
- Schemas too large for the context window, causing partial or hallucinated structure.
- No versioning when schema changes break saved drafts or integrations.
Frequently asked questions
What is Schema?
A schema defines the shape of data: required fields, types, allowed values, and relationships between properties.
Why should designers care about Schema?
Without a shared schema, teams get brittle prompts and broken UI when the model omits fields, renames keys, or drifts format across releases.
What are common mistakes with Schema?
Letting the model invent field names per request with no stable schema; Schemas too large for the context window, causing partial or hallucinated structure; No versioning when schema changes break saved drafts or integrations.