A human-machine verification method and apparatus, a device and a storage medium, relating to the fields of Internet security technologies and computer vision technologies are disclosed. An implementation includes: receiving an identity verification request sent by a requester, and collecting an eyeball gaze point track on an identity verification page, the identity verification request including identity verification information; identifying whether the identity verification information is correct based on pre-stored user identity information; if the identity verification information is correct, performing classification based on the eyeball gaze point track on the identity verification page using a first classification model, and outputting a first probability value indicating whether the requester is a real person or a machine; and determining based on the first probability value that the requester is the real person or the machine, and outputting an identity verification result indicating that identity verification is passed or fails.