One question, two answers. A model answers from memory — just as confident when it's wrong. The grounded one cites the record, or hits a dry hole.
Grounded through Drillable+ verified source
A model on its ownno source
Often both are right — a frontier model knows most facts. The point isn't who's right: one answer cites a source you can drill, flags when sources disagree, and will say “no record” rather than invent. That's what the corpus is for. Try to break it.