The point of the summary is also to explain "why" something was done, most Claude-generated PR descriptions I've been seeing go through the "what" and "how" but if the human-in-the-loop didn't care to precisely describe the "why" it is just an English version of the changes made in the code... I can just read the code for that, give me the reasons behind the diff and I'm a happy camper.

If you have a large PR the existence of a good summary on "what" changed can help you to make a better review.

But I agree with you, when reading PR descriptions and code comments I want a "why" not a "what". And that is why I think most LLM-generated documentation is bad.