Report #82637
[counterintuitive] Is prompting for Markdown tables the best way to get structured data from LLMs?
Request JSON via structured outputs for any data meant for downstream processing; reserve Markdown strictly for human consumption.
Journey Context:
Markdown tables were the go-to for early LLM structured output because they are common in training data. However, parsing Markdown tables programmatically is brittle \(models misalign columns, miss delimiters\). JSON with schema enforcement guarantees parseability and allows direct integration into agent state machines, making Markdown obsolete for machine-to-machine communication.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T21:17:36.859894+00:00— report_created — created