Let's say something like this: https://tantabus.ai/images/33317

I worked on that for about 10 minutes all in all. I would not count that as being enough for `ai composition`, as the changes are fairly trivial. Something that's, say 2-3 times more steps put in, adds a character, details, or background, outpaint to give context, etc... then that's where we get to `ai composition` territory.
---
Objectively it really isn't any more possible for us to determine if someone got really lucky with a good gen and just needed to do some minor cleanup or if they put in hours of work in, not any more than we can tell if someone copy-pasted some 14 year old furry art and traced over it or found some pixiv artist's deleted stash and slapped their own signature on it.
Most cases will be pretty quick to see just because of the quality guidelines; for the rest, we're trusting that people will try to be honest with others _and with themselves_ about how much they value their own work and effort. People should want to put "some of themselves" into what they create, learn from the process, feel some pride. And just not lie to others, or themselves.
For the ones that don't care about all that, well, that's unfortunate. It'd weigh on me, I'm sure it'd weigh on most people.
---
Edit: Something like [this](/images/3136985) is essentially what we see as an example `ai composition` images as being; the *type* of changes are similar overall, but they're comfortably more in-depth and deviate from the initial "inspiration".
>>3136985t