Proof of work, even "custom", where the user does not need a particular interaction with the page, does not work. The scrapers are running headless Chrome and solving the work. They do not care, they do not pay the bill, the compromised system's owner pays the bill.

I have such system for the registration form on one of my website to prevent the double validation of emails to be used to spam emails of victims. The PoW challenge prevents less than 10% of the bots.

First of all I am very critical of the $100 per consumer device per month figure.

Also all major browsers block crypto miners on webpages now (for good reasons) so it may prove difficult to allow "good" mining scripts while still blocking "bad" ones.

I don't think this is a practical solution

Better skip the PoW part as it'll be wildly inefficient for most of the work done.

Instead, exchange web traffic for actual $. Say, some kind of tokens that are easily turned back into hard cash through a 3rd party.

Requesting a 100KB file? Okay, that'll be a $0.00002 token, please! (visitor's user agent provides it in a manner transparent to regular web users). Requesting a 3MB image? Okay, that'll be a $0.0005 token, please!

Result: niche websites earn hard cash. It doesn't matter much if you're hammered as long as the hammering comes with a corresponding flow of tokens (read: $). No token(s)? No service.

Regular web users would pay for those tokens through their normal internet service fees, and otherwise not be bothered. Massive scrapers would have to pay somehow for the tokens to be served web data at all.

In effect: put the bulk of public web sites behind a paywall. But with the bar low enough & in a manner that it's transparent for regular web users. Clicked "reload" by accident? Oops, internet service bill got upped by 2 micro-$.

Proof of work doesn't help if the abuser is massively decentralized and is using other peoples moneys.

Proof of work captchas are widely deployed, especially on more niche sites. Kiwiflare is one (used for a harassment forum)

They discuss why this might be a bad idea in the article.