> The only prudent approach is to assume all results are bad until proven otherwise.

As is true with results from people.

> But this approach negates much of the incentive to pay for questionable results.

I'm not sure that follows. Even the crack-addled human informant has always been paid for questionable results.