Critics, recommendations from existing known authors, publishers being clear about what is and is not ai generated.

This. Ironically, an increase in slop production could end up making it harder for new non-AI authors to get recognized, due to default assumptions that unknown quantities are slop.

>> mostly of low effort copied games

All AI games are heavily bloated. Try cloning Factorio or Prison Architect. AI can make something that looks similar to such games, but the back end falls apart.

I have friends who keep showing me their new "games". I just see some version of Doom, but one that needs multiple GB of memory to accomplish what was commonplace in the 90s with a millionth as much headroom.

If you want to judge more entries without increasing the work for the judges, you need to scale the number of judges with the number of entrants, i.e. use peer review. Of course if most entries are low-effort, you'll get a lot of low-effort reviews. But if you have an inner circle of trusted judges, you can have them look at the winners according to peer review and, if they don't meet the bar, throw out reviewers who rated them highly. Repeat until you've found a subset of reviewers who are taking the task seriously.

> a legit writer with something real to say

What does that mean? What defines a "legit writer" and who determines when they're saying "something real"?

you increase the requirements.

Steam wants 100$ from you, tax documents etc.

After that, you make the initial review very cheap for you: YOu need to create a promovideo from ingame.

If this is still too much, you make a pre sampling phase. Everyone already developed a game gets more karma points and is easier found, everyone new has to earn their standing.