Why not advertise based on what they consumed last week too? And once you start down that path it's like the Netflix recommendation problem. Also, at scale one Hope's they refine the algorithms based on actual CTR feedback.
from my personal experience because it takes a way too static view of my consumption habits. Echoing some other replies here I find these personal recommender systems to be completely awful.
They constantly recommend me stuff I've already bought or have no interest in any more, and Netflix constantly overfits what I like, vastly underestimating genres I haven't seen yet while recommending me awful shows presumably just because they fall into some sort of similar buckets to higher quality shows I've watched.
Case in point, the NYT already got rid of behavioural targetting in Europe last year and actually saw ad-revenue go up. (https://techcrunch.com/2019/01/20/dont-be-creepy/)
Maybe you're thinking of how it is now, but when Netflix had its sophisticated Cinematch, its recommendations were much, much better.
In similar vein, why not just advertise based on the content the person is currently consuming? Put fishing ads into an article about fishing, the same way people put gun ads into gun magazines.
Given how abysmal targetted advertising is I don't see this being much worse.