What This Looks Like
The AI returns two or more fields, labels, keys, columns, or sections that appear to mean the same thing. The names may differ slightly, but the values overlap enough that the user or downstream system cannot tell which field is authoritative.
Why It Matters
Duplicate fields create ambiguity. A parser may accept the output, but reviewers or downstream systems still have to decide which value to trust. If the duplicate fields later diverge, the output can carry conflicting meaning even though each field looks valid on its own.
Structural Signal
The output contains redundant declarations in the same structural space. The issue is not that extra prose was added; it is that the structure contains multiple fields or labels competing to represent the same meaning.
Common Triggers
- The prompt uses multiple terms for the same concept
- Examples include overlapping field names
- The model tries to preserve both user language and schema language
- A schema migration leaves old and new names in context
- The AI creates helpful-looking aliases as separate fields
- The output contract does not declare one canonical field name
When to Use This Issue
Use this Issue when structured output contains duplicate or overlapping fields that create uncertainty about which field should be read, stored, validated, or passed forward.
When Not to Use This Issue
Do not use this Issue when fields are merely related but carry distinct meanings. Do not use it for repeated prose unless the repetition creates a structural ambiguity in the output.