I firmly believe (as in assuming, no proof) that it was both a cost saving measure and what you describe as increasing revenue - think about it in terms of stat

I firmly believe (as in assuming, no proof) that it was both a cost saving measure and what you describe as increasing revenue - think about it in terms of statistics, if you give the user the most relevant videos as the top 5 suggestions, you can cut down the cost of a query substantially and so cheaply so that you spend micro-cents per query, whereas before you'd arbitrarily cut off search results related to the user query after ~50 pages or potentially more, costing maybe thousanths-cents per query The new scenario gets you the user conversion to a watched video and subsequently to a relevant ad, all the way to the 95th percentile, and maybe even 99th percentile conversion, at a fraction of the cost, and doubled with the fact that the extremely relevant result got the user to a video, got more monetization as a result. Your cost is then simply the query time to the cheap edge model, and then the search that provides the most relevant videos to search top-n (in this case 5) It's not working for some people - but who cares, you hit 99th percentile of all users and you've cut down the spending of a black hole that YouTube is, in fact you might've even made the experience better for 99% of users. I have nothing to substantiate my claims, but knowing how I'd reason about it, this would be my assumption for the state YouTube search is in I also hate this and it has made me move away from YouTube search, unfortunately YouTube recommended is quite good at giving me relevant videos to myself