Although acupuncture is an effective therapeutic intervention for pain reduction, the exact difference between real and sham acupuncture has not been clearly understood because a somatosensory tactile component is commonly included in the existing sham acupuncture protocols. In an event-related fMRI experiment, we implemented a novel form of sham acupuncture, phantom acupuncture, that reproduces the acupuncture needling procedure without somatosensory tactile stimulation while maintaining the credibility of the acupuncture treatment context. Fifty-six non-specific low back pain patients received either real (REAL) or phantom (PHNT) acupuncture stimulation in a parallel group study. The REAL group exhibited greater activation in the posterior insula and anterior cingulate cortex, reflecting the needling-specific components of acupuncture. We demonstrated that PHNT could be delivered credibly. Interestingly, the PHNT-credible group exhibited bilateral activation in SI/SII and also reported vicarious acupuncture sensations without needling stimulation. The PHNT group showed greater activation in the bilateral dorsolateral/ventrolateral prefrontal cortex (dlPFC/vlPFC). Moreover, the PHNT group exhibited significant pain reduction, with a significant correlation between the subjective fMRI signal in the right dlPFC/vlPFC and a score assessing belief in acupuncture effectiveness. These results support an expectation-related placebo analgesic effect on subjective pain intensity ratings, possibly mediated by right prefrontal cortex activity.