Big sites like Facebook and Reddit can avoid using them because they have teams dedicated to identifying bot accounts algorithmically. But anyways those aren’t working well so those platforms and their users are suffering as a result.
Simply from personal experience. The majority of the time when I hit a captcha or somesuch, it's been for websites of marginal usefulness to me.
It's certainly not some sort of hard and fast rule -- there are plenty of exceptions as you point out. That's why I consider it a weak signal. But if other weak signals make me doubt the site is useful and I hit a captcha, I'll just move on immediately rather then continuing to spend time checking it out.
I think it's mostly that doing a captcha is a cost to me in time and irritation, and if I don't think the website is worth paying that cost, I won't.
I've come to view the existence of these checks at all as a weak signal that the website in question is probably not worth my time.