There are many startups attempting to build content recommendation and curation algorithms. Their approches are varied and range from the simple to the very complex.
I would generally categorize the simple approach as using explicit profiles. Explicit profiles are built by asking directed questions (What are you interests?) and tracking usage (What articles have you read?)
The complex approach is to build implicit profiles. Implicit profiles are based on what you do, but also what you don’t do. This requires a lot more understanding of the content characteristics and mapping them back to the user profile. For example, you have be shown 10 different articles about a specific topic, but you only click on 1 of them. What was the reason you clicked on that specific article? What it the image? author? title? time of day? day of week? device (mobile/tablet/desktop)? What if you already read it from a different source?
Recommendation algorithms are designed to be safe. If you’re on Amazon looking at a book, you won’t be recommend a scooter. It will most likely recommend another book within the same genre and topic. This bodes well for a the explicit profile approach to work without the need a very strong implicit profile.
There are many other issues to overcome with content recommendations and curation, I don’t think implicit profiles will be the main hurdle.