This is entirely anecdotal but recently many top results in DDG read like they’re written by GPT-3 or the like. Half coherent sprawls of text that contain lots of random factoids but nothing novel or that actually answers the question.