Blog
How to test an LLM search summary for invented results
6 September 2026
A runnable Node.js example for rejecting unseen project IDs, preventing invented display names, and understanding what passing tests cannot prove.
Our LLM search feature fabricated results — the guardrail said "never fabricate IDs"
8 July 2026
A production debugging story: why an AI summary invented search results that didn't exist, why the obvious fix makes it worse, and the three-layer fix that shipped.