Time series CCD photometry of the inner coma of active comets has been used over the last twenty years to determine the rotation period of their nucleus. Usually, the photometry is performed by using very small apertures with the aim of isolating as much as possible the brightness contribution of the nucleus with respect to the brightness of the coma or of measuring short-term gas/dust production rate diurnal variations. The effect of seeing variations in the brightness profile of the comet and their consequences on the measured magnitude in such small apertures has never been studied before. The aim of this paper is to study this ``seeing effect.'' A series of simulated images of two comets is generated by using different seeing values. The brightness of the simulated comets is measured by using very small apertures ranging from 2 to 16 pixels. A strong correlation between the measured magnitudes and the seeing is found. This systematic effect can produce false variations of several tenths of magnitudes in the light curve in nights with variable seeing conditions. A new method to correct this seeing effect on real observations is proposed and applied to CCD observations of comets 29P/Schwassmann-Wachmann 1 and 21P/Giacobini-Zinner, carried out with the 82 cm IAC-80 Telescope at Teide Observatory (Canary Islands, Spain). The seeing effect is clearly present in the observations, and the method is shown to be very efficient in correcting it.