In general: attention of any person is limited. To monetize attention, it is necessary to steal it from things the user want to pay it to.
> Reducing friction for many sites means that users will spend more time browsing on the site.
The users usually want to spend as little time as possible on any given site. They're using it to accomplish some goal - like find a piece of information, or pick and buy an item, or be entertained. Even with YouTube: autoplay may be something the users want, and in this case it reduces friction, but money is made on the users being exposed to ads - ads before, after, during and surrounding the video. All these are friction, taking both time and attention away from watching the video itself. This is where YouTube makes money.
Showing ads is an obvious case of introducing friction to monetize attention. There are subtler approaches too. They're insidious and so pervasive that some became examples of "good design" on the web. There's too many to list them all. For example, think of all the cases where you thought, "this UI is dumb" or "this design is inefficient". Chances are, it's because it ensures you have to stay on the page longer, click around more, possibly get confused.
I'll give one specific family of examples: a UI/UX pattern in e-commerce, where each item is featured as a large tile or "media item" on a list. A big picture, a name, a price, perhaps one or two pieces of detail. Only 4-6 results fit on the screen at any single time - where a better design could make the site fit 20+ items instead[0]. You start clicking on them individually, and notice each item has a different set of details, possibly using different units - making it impossible to compare options. That's not accidental. That's designed to frustrate user's ability to compare items and make a good choice, in hopes they'll just stick to whatever the store surfaces at the top.
--
[0] - I know because I did that. Last time I was assembling my PC, I spent an hour or two writing custom CSS rules for a major Polish electronics retailer, and was able to turn their 4-5 items/screen into some 30 items/screen. Without this, trying to pick among many similar options was too much cognitive burden for people to bother. As intended.