They do though. The point of captchas isn’t to block all bots, that’s never going to happen. They are there to block the low effort bot army that’s rampaging across the internet which is effectively a DDOS stack on smaller servers.

Big AI companies throw petabytes of Data into their models. Instead of crawling and saving it once, they just index each site worth using and scrap the data each time they train a model. As a server owner you either block the scrappers with a captcha or you blanked ban IP ranges that are known for scraping.