TikTok, more than any other company, is built on it's recommender. Here's how it works (sort of):

A billion people spend time on the video sharing app TikTok every day. It offers an endless stream of short videos and it tailors itself to your interests. Did you enjoy that video of a dog carrying a big stick? Here's some more! Like that list of top 5 council estate lads' names? Here's a picture of a lame UK police chase! Magic!

How does this all work? The New York Times recently acquired a document showing some of the details. They are not particularly surprising. The site optimises for two main things "retention" - whether a user comes back - and "time spent". Like all similar sites: it's aiming to get you hooked and ramp up the dose. Behavioural inputs include likes, comments and playtime. There's not much on the nitty gritty detail, but an interesting read anyway.

I think the main thing that makes the algorithm work is that it's easy to collect lots of data on a user quickly - videos are short and all share the same format.