The OP seems to be coming from the perspective of "my time as a PR reviewer is limited and valuable, so I don't want to spend it coaching an AI agent or a thin human interface to an AI agent". From that perspective, it makes perfect sense to want to know how much a human is actually in the loop for a given PR. If the PR is good enough to not need much review then whether AI wrote it is less important.
An angle not mentioned in the OP is copyright - depending on your jurisdiction, AI-generated text can't be copyrighted, which could call into question whether you can enforce your open source license anymore if the majority of the codebase was AI-generated with little human intervention.