Self-driving car fleets probably increase suburbization in the same way that trains created suburbs long ago.
Depending on how it breaks out, "more space per person at similar total environmental impact" is a quality of life win but not a climate-change-fighting one.
Pedestrian safety and other externalities are probably a big win, still.
The parking lot point is interesting, I hadn't thought about that.
But it's not obviously better to me: commuting behavior is very time-of-day determined, and the cars are gonna have to be somewhere in between rush hours. You don't want them on the road, and you can probably shift around to catch other sorts of predictable demand, but environmentally you're gonna want to minimize the total nubmer of miles driven without passengers.
EDIT: so in areas with less development and commmuters are parking their cars in surface lots all day, yeah, you don't need those surface lots for commuters + all the other lots at retail, residential, etc all at once if the cars are moving around to other uses during the day. But in denser recent-development areas where office parks predominantly have underground or garage parking it may be harder to realize infill there.