This paper proposes a time‐domain fitting method for on‐site calibration of the air‐coil sensor. The air‐coil sensor has been widely used in transient electromagnetic exploration. Due to limited bandwidth of the coil, the output signal is distorted, causing a phenomenon known as the transition process. To accurately measure the magnetic field from the output signals, the relationship between the coil induced electromotive force and the output voltage must be confirmed by on‐site calibration, which requires high calibration accuracy and demands simple operation, portable equipment, and adaptability to the environment. Conventional frequency response methods, however, requires a uniform magnetic field with various frequencies to obtain the frequency response curve of the air‐coil sensor. The time to acquire the signal correlates with the number of test frequencies, and the equipment used to generate a uniform magnetic field must be tailored to the shape of the air‐coil sensor under test. This paper constructs a relationship between the calibration file and the zero‐input response of the air‐coil sensor and designs an optimization algorithm to suppress the soil eddy current effect. This on‐site calibration method lifts the dependence on the uniform calibration field and reduces significantly the time required for calibration. The calibration source can be generated by cutting off the voltage source in parallel to the calibration coil, which greatly reduces the cost of the signal generator and provides a better solution for realizing the embedded self‐test devices. Experimental results show that the proposed method effectively improves the calculation accuracy of the apparent resistivity.