The problem is that SEO outfits gamified that metric as well. The idea of determining whether a site is useful based on how long the user stayed on it only served to cause sites to hide information halfway down the page.
That idea is THE CAUSE of the meme about recipe sites telling you how your long-dead great-great-great grandma wrote this recipe for a special event and how it has become a family staple and some of the ingredients might not exist anymore in the same way they did, before they get around to giving you the recipe.
Recipe sites were HUGE in the heyday of the DotCom boom and the only way to get your page to rank higher than every other online cookbook was to capture people's attention with a story before showing them what they came there for. If they can just get what they came for and leave, you got downranked as not a useful site.
Oh. Um. Well, hypothetically, maybe they do but the metric they're optimizing is their profit instead of user-facing quality?
I just assumed that everyone already knew this. Of course Google could fix their results. They want them this way.
Read the emails released in the recent antitrust suit and you'll see intense pressure not to make changes to search, or even to Chrome flows that drive users to Google search results.
Bottom line, if ads is performing, what incentive is there to change?
The consensus seems to be that SEO has won, that experts are stuffing websites with keywords and the search engines are simply unable to figure out what is good content anymore.
So why not use your own browser to measure how people interact with a website? How long do they stay on it? Do they just scroll and exit immediately? Do they seem to have a hard time finding information? Maybe randomly ask people if a website is helpful or not from time to time. Then aggregate that info and score a website. This should cut down on website that have nothing but fluff.