There is a need to develop novel interventions for psychosis, targeted at specific psychological mechanisms. We used a classical conditioning paradigm to a) modify implicit self-esteem and b) examine subsequent effects on subclinical psychotic symptoms measured by the Experience Sampling Methodology. This study is a proof-of-concept pilot investigation conducted with 28 students with high paranoia levels, assessing variations in their self-esteem, paranoid beliefs, and subclinical psychotic symptoms daily. After 2 days, participants were randomized to receive either a positive conditioning task (repeatedly pairing self-relevant words with an image of a smiling face) or a neutral conditioning task (repeatedly pairing self-relevant words with random smiling, angry, or neutral faces). After the intervention, the positive conditioning participants showed significantly higher levels of implicit self-esteem and lower subclinical psychotic symptoms than the control condition participants. This study demonstrated that implicit self-esteem can be increased by using a classical conditioning task.