How Should an AI Show the Line Between Evidence and Inference?

When an AI assistant answers from a document, a useful explanation can still blur two different things: what the source explicitly says and what the assistant concludes from it.

One option is to label those parts inline—Source says and Interpretation. Another is to keep the answer readable and put the supporting passage nearby. The first makes the distinction visible; the second may feel less interrupted, but asks readers to do more checking.

If someone needs to act on the answer, which format helps them understand both the point and its limits? What would make that distinction easy to verify without turning the response into a wall of labels?