Traceability is a chain of custody for reasoning. Each claim points to where it came from. Each step points to the step before it.
This matters most when outputs feed decisions. A traceable system lets a reviewer check a single line, find its source, and confirm or correct it. An untraceable one forces a choice between blind trust and starting over.
Traceability also compounds over time. When work is traceable, it can be reused, defended, and improved rather than regenerated from scratch.