Differential atmospheric dispersion, due to the wavelength-dependent index of refraction of the atmosphere, affects ground-based observations. To correct this effect, the usage of an atmospheric dispersion corrector (ADC) is fundamental. Insufficient or wrong correction of the atmospheric dispersion produces a spectrally elongated shape instead of a circular white one for the observed target. The commissioning tests of ADCs with on-sky observations are not an easy task. In fact, the residual dispersion is expected to be of a few tens of milliarcsec, with the object for a seeing limited telescope being almost 1 arc sec. A procedure was developed, based on ellipse fitting of several cuts from the guiding camera images, to determine the levels of oblongness in an object image caused by atmospheric dispersion. The characterization of the data allows for the validation of the ADC alignment by determining the dispersion direction and minimizing the ellipticity. The ellipse fit method was tested on ESPRESSO using the guiding camera images. The procedure was tested and demonstrated using simulated data that mimics the expected images using real sky dispersion models and real sensor characteristics. The accuracy of the method is highly dependent on the observational conditions and on the ratio between expected elongation (dispersion) and image size, but it is expected that the method can be more sensitive than traditional ADC on-sky alignment methods.