Making people go through hoops will just discourage legitimate potential contributors and not stop AI slop. LLMs are good at generating legitimate sounding wall of text. Without actual code, it'll be harder to distinguish legitimate contributors from spammers.
You could ask the submitter to show a quick video recording of the new feature being used. Or if its a bugfix, show the failure scenario and then the fixed non-buggy scenario. If they can't be bothered to show a basic before/after demo of whatever they are working on, then you probably don't want to work with them and accept their code changes anyway.
People already go through hoops and live with it just fine. I don't claim to have the best solution but fundamentally its a social problem and therefore solvable. Perhaps some form of chain of trust.