Objective. Neural mass model (NMM) has been widely used to investigate the neurophysiological mechanisms of anesthetic drugs induced general anesthesia (GA). However, whether the parameters of NMM could track the effects of anesthesia still unknown. Approach. We proposed using the cortical NMM (CNMM) to infer the potential neurophysiological mechanism of three different anesthetic drugs (i.e. propofol, sevoflurane, and (S)-ketamine) induced GA, and we employed unscented Kalman filter (UKF) to track any change in raw electroencephalography (rEEG) in frontal area during GA. We did this by estimating the parameters of population gain [i.e. excitatory/inhibitory postsynaptic potential (EPSP/IPSP, i.e. parameter A/B in CNMM) and the time constant rate of EPSP/IPSP (i.e. parameter a/b in CNMM). We compared the rEEG and simulated EEG (sEEG) from the perspective of spectrum, phase-amplitude coupling (PAC), and permutation entropy (PE). Main results. Under three estimated parameters (i.e. A, B, and a for propofol/sevoflurane or b for (S)-ketamine), the rEEG and sEEG had similar waveforms, time-frequency spectra, and PAC patterns during GA for the three drugs. The PE curves derived from rEEG and sEEG had high correlation coefficients (propofol: 0.97 ± 0.03, sevoflurane: 0.96 ± 0.03, (S)-ketamine: 0.98 ± 0.02) and coefficients of determination (R 2) (propofol: 0.86 ± 0.03, sevoflurane: 0.68 ± 0.30, (S)-ketamine: 0.70 ± 0.18). Except for parameter A for sevoflurane, the estimated parameters for each drug in CNMM can differentiate wakefulness and non-wakefulness states. Compared with the simulation of three estimated parameters, the UKF-based CNMM had lower tracking accuracy under the simulation of four estimated parameters (i.e. A, B, a, and b) for three drugs. Significance. The results demonstrate that a combination of CNMM and UKF could track the neural activities during GA. The EPSP/IPSP and their time constant rate can interpret the anesthetic drug’s effect on the brain, and can be used as a new index for depth of anesthesia monitoring.