I don't think it is a false dilemma. The problem is about the difference between the magnitude of revenue generated by the proposed methods. Both for publishers

I don't think it is a false dilemma. The problem is about the difference between the magnitude of revenue generated by the proposed methods. Both for publishers AND people interested in advertising products. Tasteful banners are nice but tracking the performance of such ads is very hard. Ad business is really cut-throat. When you visit a site, there literally are multiple ad networks / advertisers bidding for a spot in front of you, with real money, based on the site, your location and your interests. A huge network of autonomous trader bots that trade real money billions of times per day for a spot in front of a human's eyeballs. If an algorithm decides that you have a higher than normal probability of converting to a paying customer, they (an AI or expert system) decides on the spot to pay "premium" to be in front of your eyeballs. They do this while running A/B tests with different ad variations to figure out what converts better, for which segment of the population etc. Modern ads are not only kind of targeted. They are INSANELY targeted. There have been instances of people targeting singular people through ads based on geolocation and interests. A publisher cannot be expected to run all that bidding code on their own servers. Those networks are actually a network of networks. If you are buying ads today, you want analytics immediately to see how you are doing, how much you are spending so that you can improve and adapt before sinking more money on it. The profits generated by such methods are a couple orders of magnitude higher than a simple untrackable banner on a site placed by a site owner which gets swapped out whenever...