I mean, theoretically, a robot would easily be able to mimic a human-like mouse motion to a Google recaptcha box. But here’s my counter-question: would it be efficient enough to justify it’s use?
Bots (software bots) that solve captchas are employed in pretty huge quantities from what I understand. Imagine having a big room full of robot hands solving captchas for people. It’d probably be easier for someone looking to circumvent them to use the existing captcha-solving software (that has become good enough to outsmart most of them.) It’d be less expensive and you’d get the work done more quickly.
That’s my thinking, at least.