I think better calibrated experts use LLMs in e.g. adjacencies to their professional domain fairly pervasively these days, accepting some risk of a hallucination based on e.g. cost of checking evidentiary trail, how important that output is and where it is exposed, etc.