You 'can' use an MSK demodulator for demodulation of a GMSK signal but it will sub-optimal. A Gaussian filter needs to be there at the receiver for matched filtering.
You can either run Monte-Carlo simulations to asses the BER penalty v/s reduction in complexity obtained by using sub-optimal MSK receivers or you can try to search for existing papers on the topic.
I've run into sub-optimal OQPSK receivers being used for SOQPSK signals several times and I suppose the same theory should apply for GMSK as well.
