BACKGROUND: Concurrent transcranial magnetic stimulation and electroencephalography (TMS-EEG) is an emerging method for studying cortical network properties. However, various artifacts affect measurement of TMS-evoked cortical potentials (TEPs), especially within 30 ms of stimulation. OBJECTIVE/HYPOTHESIS: The aim of this study was to assess the origin and recovery of short-latency TMS-EEG artifacts (<30 ms) using different stimulators and under different experimental conditions. METHODS: EEG was recorded during TMS delivered to a phantom head (melon) and 12 healthy volunteers with different TMS machines, at different scalp positions, at different TMS intensities, and following paired-pulse TMS. Recovery from the TMS artifact and other short-latency artifacts were compared between conditions. RESULTS: Following phantom stimulation, the artifact resulting from different TMS machines (Magstim 200, Magventure MagPro R30 and X100) and pulse shapes (monophasic and biphasic) resulted in different artifact profiles. After accounting for differences between machines, TMS artifacts recovered within ∼12 ms. This was replicated in human participants, however a large secondary artifact (peaks at 5 and 10 ms) became prominent following stimulation over lateral scalp positions, which only recovered after ∼25-40 ms. Increasing TMS intensity increased secondary artifact amplitude over both motor and prefrontal cortex. There was no consistent modulation of the secondary artifact following inhibitory paired-pulse TMS (interstimulus interval = 100 ms) over motor cortex. CONCLUSIONS: The secondary artifact observed in humans is consistent with activation of scalp muscles following TMS. TEPs can be recorded within a short period of time (10-12 ms) following TMS, however measures must be taken to avoid muscle stimulation.