Suppression of spinal responses to noxious stimulation has been detected using spinal fMRI during placebo analgesia, which is therefore increasingly considered a phenomenon caused by descending inhibition of spinal activity. However, spinal fMRI is technically challenging and prone to false-positive results. Here we recorded laser-evoked potentials (LEPs) during placebo analgesia in humans. LEPs allow neural activity to be measured directly and with high enough temporal resolution to capture the sequence of cortical areas activated by nociceptive stimuli. If placebo analgesia is mediated by inhibition at spinal level, this would result in a general suppression of LEPs rather than in a selective reduction of their late components. LEPs and subjective pain ratings were obtained in two groups of healthy volunteers - one was conditioned for placebo analgesia while the other served as unconditioned control. Laser stimuli at three suprathreshold energies were delivered to the right hand dorsum. Placebo analgesia was associated with a significant reduction of the amplitude of the late P2 component. In contrast, the early N1 component, reflecting the arrival of the nociceptive input to the primary somatosensory cortex (SI), was only affected by stimulus energy. This selective suppression of late LEPs indicates that placebo analgesia is mediated by direct intracortical modulation rather than inhibition of the nociceptive input at spinal level. The observed cortical modulation occurs after the responses elicited by the nociceptive stimulus in the SI, suggesting that higher order sensory processes are modulated during placebo analgesia.