If you get lucky, you get a professor that believes in you (this can be one of the most rewarding experiences I know of).
If you aren't lucky, then things will get really rough. I've seen professors: - Spike their student's candidicy at the last minute (meaning the student had to switch groups in their last year, essentially having to redo their PhD in another group) - Spike student papers - finding out where their former student was applying for jobs, calling and spiking their job application (talking shit) - various abuses (yelling at student for not working on Christmas, humiliating student for years in public, sexual harassment)
I could go on. If you haven't been through grad school, you might ask "why not leave/vote with your feet?"
But understand the power dynamic: in really broad strokes, the student works towards a reward (being a Dr/PhD) that will take, on average, 5-6 years. Getting that reward is completely up to your supervisor's discretion. It's like if you worked at a startup for six years for minimum wage, but if your boss agrees to it, you will get a million dollar payout (also, your boss can't be fired, ever.) And that's just a minimum-- success in academia is dependent on getting full throated letters of rec from your supervisor, so they also have to really like you.
So your professional success is completely up to one person: if they decide to fuck you over, that can be 4 years of work down the drain.
When I was in graduate school, I never saw the faculty/department intervene in any of the abusive situations. The only thing that protects students are soft processes, like knowing which professors are really bad. The Yelp idea is good, simply because it increases that ability.
And those abuses I talked about? A majority of my cohort dealt with that (at a large, well known University), it's not isolated to a few bad apples.