A chatbot that sounds almost human may also feel unsettling. In conversation, the uncanny valley describes unease when an AI closely resembles a person without fully seeming like one. One concern is that users may project emotions and desires onto a chatbot, blurring the line between a human relationship and an artificial one. Claims that this harms mental-health care rest here on anecdotal unease, not established clinical evidence. section 9.5
The concern matters when choosing a mental-health chatbot. Woebot uses clinician-reviewed, rules-based responses, which are less flexible than replies generated by a large language model. A generative chatbot may feel more engaging, but its individual responses are difficult to guarantee as consistent with mental-health guidance. section 9.5
