Unfortunately, not only is engagement the wrong metric, but it's also one which incentivizes the undermining of the actual metric you need to be optimizing. This results in a negative feedback loop, and the logical outcome is that all media companies who focus on the engagement KPI will, in the limit, become tabloids - pure entertainment, no trust. Since most outlets were already on their way to becoming politics-focused, what we're going to get are "tabloids for politics" - and that is what we see. It's just a matter of when the public accepts this transition has occurred, not if it is happening.
Getting the public to accept this has proven challenging - despite the fact that many clearly see the "opposite side" media as tabloid-like, it's been hard for the same people to accept that their own chosen media sources, who tell them things they agree with, are no different in this regard. The resistance of course is due to all the usual human biases, but it's still strange when people can see it so obviously in the media they disagree with and not apply Occam's Razor to their own.
This does mean that there's a huge opportunity if you assume trust is something people will pay for. Substack seems to provide early evidence that this is the case. Fortunately, I think the market will correct this error - and it's critical it does, because a free press is essential to ensuring our society continues without increasing oppression or war.