There's a positive feedback loop involved in any transit investment and how it affects the neighborhood, but especially so with rail - unquestionably so as you think at bigger scales. One of the touted benefits of California High Speed Rail, for example, is that it would bring the Central Valley cities into the commuting sphere of the other urban centers.
(The worst thing you could do with a rail stop, from a land-use perspective, is put a big surface parking lot around it. That turns it into a transfer point for a car commute, not a prestigious destination that other transit will naturally run toward. One of the problems with the CHSR plan, the last time I looked, is that the Central Valley alignments are a pretty dodgy package.)
That said, providing better transit at lower cost is always an upside, since the first-order question for trip planning by transit is always the overall level of service - frequency and coverage, as well as speed. I would expect robobuses to usher in a new era for system planners.
So even if it's more expensive upfront after X years you get continuous savings every year from increased efficiency.
When you go rail you immediately restrict yourself to that route. There is little to not flexibility. Let alone rails on the roads in snowy climates just cause additional issues. As for simulating with buses, you can embed in the roadway or even on light poles or similar along the route where you want the bus to go when automated. Hence its "on rails" without physical rails.