There are several hurdles to getting users to publish consumable content with your tool:
- They have to be able to afford it
- They have to have the skills to meet your software's easy to use
- They have to want to share the content they've made
- Others must want to consume the content they've made.
That leaves us with the crunch - 4 - creating content that the world wants to consume. It's the step where traditional apps fall on their arses. MS Word content consumable? Give Photoshop to the man on the street and they'll draw a penis on their holiday snaps and forget about it.
This is the web's approach - a Million Monkeys at a Million Keyboards and let the stuff that isn't pictures of penises float to the top using social/viral techniques. This is good for the talented (and lucky), but for the rest of us we end up with drivel and don't want to publish. This is lost content and lost revenue.
This is where AI comes into it. A system where it is harder to create drivel than to create consumable content lands you squarely on the UGC bandwagon. Guaranteed worst case user generated content. This is currently the realm of games like Spore's creature creator, that use machine intelligence* to let people make and publish consumable content.

It is hard to create bad content with the creature creator, and it is a hit with 39,502,095 creatures posted to the web as I'm writing this. The only thing wrong with the system was that that EA has failed to monetize it's huge library of content...
*So one good way of introducing my subject (procedural geometry) is as a tool to create guaranteed worst case UGC. There are other reasons described elsewhere in this blog/my old dissertation.