The issue with procedural generation is the game has to be built for it from the ground up and in a modular way. AAAs try to make themselves appealing by using novel new high quality assets that aren't modular.

I haven't played starfield so idk what they ended up doing, but from the sound of it they have pre-made assets/areas that they then place onto pre-generated worlds in a randomized way.

To make one of these "areas" procedural in itself, they'd then have to code a whole system for that. With AAA/3D the hard part is making modular environments without it looking repetitive or ugly.

My point isn't so much that it can't be done in a AAA game. But rather that it's risky to do (not all players like it), and you have to structure your development around it. Lots can go wrong, there's stuff you gotta sacrifice to make it work, etc.

If starfield is on the old bethesda engine then that's even more of a reason. You can't just plug and play an entire procedural generation thing in there without some fairly large overhauls or just gluing on an unrelated system.

In practice, bethesda probably took the lazy route: using their existing engine without major changes, then just making new assets for it, throwing stuff about a bit randomly, and calling it a day.

That's the thing about procedural generation is: it's a lot of effort and sucks up a huge part of the game's development and comes at some pretty strict costs (repetitive looking environments/gameplay, reduced novelty, larger programming dev time to make it work). It can be done, but for a cost-cutting AAA studio they're not gonna bother.

yes. I haven't played the game so idk the details of what's up. but at 1k+ planet-sized spaces it's hard to have a team go over that by hand. Planets are large. But I have no doubt that bethesda team was probably super lazy as well.