How do I handle malformed JSON from an LLM without retrying?

Handle malformed LLM JSON through defensive parsing strategies. Extract JSON using regex to find content between first { and last }, handling extraneous text. Use lenient JSON parsers that fix common issues: json-repair library auto-corrects trailing commas, unquoted keys, and incomplete structures. Implement fuzzy matching to extract key-value pairs even from partially valid JSON. Parse incrementally, extracting valid portions and using defaults for missing data. Use JSON5 parsers that accept relaxed syntax. Transform malformed output through string manipulation before parsing: remove markdown, fix quotes, balance brackets. Implement fallback schemas with optional fields to handle structural variations. Log malformed responses for analysis and prompt improvement. Consider multiple parsing attempts with different strategies before failing. Validate and format recovered JSON with our JSON Formatter at jsonconsole.com/json-formatter. For critical applications, still use retries but implement these techniques as fallback. Balance robustness with correctness requirements. Prevention through better prompting remains best strategy.