I wouldn't worry a whole lot if it's been made already or not. Chances are, something fairly similar does already exist. From what I can see, most of the time, a game's success has a lot less to do with its pure originality, and way more to do with how well it's made, how fun the game mechanics are, etc, etc. A lot of the iconic and successful games out there weren't necessarily the first or most original, they just did it best. By all means, do your research and such, I'm not suggesting otherwise. But, I wouldn't stress about it too much. Just make a game you think will be fun
To your question, if it hasn't been made, offhand I would wonder if procedural generation would add all that much for a game like GTA. Generally games of that genre are rather large to begin with, and much of the gameplay is based on taking over parts of the city, doing quests, and such. Generally, I see (and use) procedural generation for games that are either quick, or let's say, repetitive, such as a strategy map (where you'll be playing repeatedly, and it avoids having to perfectly balance a map). Also, I think it's often used by small developers who don't have the time and resources to generate large amounts of content. I don't think procedural generation would take much away from your game, by any means, but I would also question how much benefit it would offer. I would worry a bit about it falling into the trying-to-do-too-many-things camp, where it wouldn't be as engaging and deep sim-city, nor as fun as gta. Though, that would be just a concern, I'm certain it's possible to pull it off.
To your other question, I don't personally know of any gta like games with procedural generation used for the game world. I'd just suggest browsing steam, itch.io, desura, etc and see what pops up.