Unverified50% confidenceFactExact time
The author summarises five common situations in production environments where LLMs return non-standard JSON: Markdown fences, polite prose prefixes, trailing commas or smart quotes, truncation, and silent garbage that is structurally correct but missing fields.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedStandardizing syntax is not the same as standardizing semantics; even if every provider returns the same JSON structure, the results are still not comparable if different accounting boundaries are used, so provenance, accounting boundary and methodology must become a core part of the discussion.65% similarUnverifiedThe most common model failure is returning something that is 'almost JSON,' such as extra commas, markdown code fences, verbose preambles, or writing the number 3 as the word 'three,' with the AI node reporting success while the workflow crashes only several steps later when parsing fails.64% similarUnverifiedQuantity assertions based on business semantics (such as validating the number of chapters) often catch truncation issues better than pure JSON format validation.64% similarUnverifiedContract drift in tool calls (such as a tool changing from returning plain text to returning JSON, or a schema adding a required field) usually does not directly produce an error pointing to the root cause, but instead manifests as the model repeatedly failing to parse, outputting garbled text, or retrying.64% similarUnverifiedghealth uses 6 exit codes (0-5) to indicate the type of success or failure, and errors are also output as JSON, so agents can branch the flow and retry based on the exit code and JSON fields.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1043342API
curl https://kongchang.com/api/v1/knowledge/claims/1043342MCP
get_claim(id=1043342)