Let users just rate that search result, so if someone else searched for the same thing and didn't find it relevant, they can thumbs down. If they found it relevant, they can thumb up.
Given how location-aware some things sometimes need to be (including spellings / alternative words, which in my experience, Google is a lot better at than DDG), I don't see how that works at all.
They will first need to rate a few hundred sites (valid ratings, because if they don't have high validity with the rest of the sites in the database the confidence this is a valid user will be low) to give the spam site a high rating. So they will have already built a database of good ratings before they can give the last vote that they get paid for?
Also, you can manually remove a spam site that uses this tactic and all their money will have been wasted.
First, each user needs to have a history or they won't be considered in the rating almost at all. Then, each user's history gets analyzed for average rating vs. the average userbase so you can determine whether this user upvotes everything or only rates the sites that are terrible.
So each user gets a score of "threshhold" of what rating site they will thumbs up. Some users will thumbs up a site that's 50% (mostly useless, but not outright spam), while some only upvote sites that are 80% (good content at the very least).
Then regressed for this variable, you can calculate their feelings on each site and see if it correlates well with other users. The more ratings, the more each additional rating is weighted.
This will reduce the number of ratings required for a statistically valid sample for each search query (since you need to tell users if it's relevant for the thing being searched, not just quality of the site). But you also need to fight bots as well.
This system may not be perfect, but it would add value to a search engine