I think you'd still get fraud, for example I'm just imagining some reddit degens sharing the comprehension answers on a wiki or discord so people can skip the time commitment of reading the thing and instead grind as many as possible with cheatsheets while watching youtube or something.
Imagine any avenue students use to cheat at classes and I think many of them apply to this system.
Each engagement required feedback, the quiz is just a layer, the mandatory feedback is the real bottleneck to get around. 3 questions about the product, and many layers of security that requires a human to have knowledge of the product, we detect any LLMs attempt to enter the fold. So even if the user tries to cheat, and he will only achieve that up to a certain point. But what the company pays for is to put their product in front of someone, and get proof it was understood many X more than a click is "understood".
In other words by a user providing feedback he is proving to a degree that he "understood" the product. That is what he gets paid for, and that is what the companies pays for.
Hot take, if you pay platforms like LinkedIn and X for your ads, they will turn a blind eye on the $63B a year ad fraud. Because they will lose billions. If you pay the "humans" directly there's no "conflict of interest". Companies get human attention, and humans get money for their time.
If you verify every human, to not be a bot, and to be reputable (e.g. employee at Google) vs. trusting LinkedIn to show your ads to those same people to want to target. The math is in the formers favor. Quality cannot be much more degraded than the current Pay-per-click solution of today, specially with LLMs crawling the web like never before.
If a human proved attention (via comprehension check, LLM detection, feedback etc.), compared to the alternative of just paying $5-$15 on LinkedIn for just a CLICK! The choice is a no-brainer.
Imagine any avenue students use to cheat at classes and I think many of them apply to this system.