I'd bet just about anything that Google uses machine learning to decide whether or not to trust a site for ads. It seems like the only solution that would work at a large enough scale to handle that kind of demand (versus more defined but more labor- and resource-intensive malware/fraud detections). I think that also explains why the review process seems so arbitrary and ineffective - in essence, not even Google know…
It doesn’t work. These automated systems are flagging a (presumably) benign site and an article yesterday regarding their $5M lawsuit for running a scam ad on their SERP for “Coinbase support” suggest the automated systems can be bypassed too.
I’m not saying automated detection can’t be a part of it, but we shouldn’t accept companies automating away decision making as if computer-derived errors are acceptable.
The larger point is that Google isn’t exactly strapped for cash. They could hire an army of reviewers. They just don’t.