I'm posting this to share the groundwork I did before applying for a PhD. It may not suit everyone, but I think it's more useful than "asking around everywhere about which research group is good."
Here's how it went. I had initially narrowed it down to two prospective advisors, but my knowledge of them was limited to those few lines of introduction on the school's official website—honestly, those few lines would make anyone's heart flutter. I didn't want to bet my next five years on a few sentences of introduction.
Later, using the talent profiles on MuchSci, I put together a "background report" for each of the two advisors. Let me talk about what I looked into.
1. Project details. This was the most informative. The projects in the profiles had undergone "exact matching of the principal investigator's name + verification of the public institution," and the project name, approval number, year, type, funding, and keywords were all attached. I focused on the trajectory of projects over the past five years—one advisor's actual projects over the past three years were clearly different in focus from the direction described on the official website. This finding directly influenced my choice.
2. Representative works. In the eight-dimension scoring, representative works carry a weight of 15. I found the representative works of both advisors and skimmed through them, which was more useful than reading ten pages of their homepage—you'll know after reading two papers whether you like the flavor of this direction.
3. Academic appointments. Public roles in societies, journals, and committees can show how active they are in the academic community and how broad their resources are. For a PhD student, this indirectly relates to how wide your "path" will be when you graduate.
4. Mentorship relationships. Publicly verifiable advising relationships. Where the students ended up is the most authentic "reference for graduation destinations," more solid than the words in admissions promotions.
The whole process took just three steps: search profiles by name and institution → go through the details and representative works item by item → form your own judgment. It was done in one afternoon.