The search for gravitational wave (GW) signals has been very successful in the first two observing runs (O1 and O2) of the LIGO and Virgo interferometers. In total, eleven GW events were detected [1]: ten from the coalescence of binary black hole systems and one from a binary neutron star system (GW170817) [2]. The latter in particular is the perfect example of the multi-messenger approach in astrophysics, thanks to the detection of an electromagnetic (EM) counterpart as a short gamma-ray burst, GRB170817A [3]. The EM follow-up performed by instruments at different wavelengths is crucial to identify the counterpart, the host galaxy, the nature of the remnant and the properties of its environment. Imaging Atmospheric Cherenkov Telescopes (IACTs) join the follow-up effort searching for very-high energy emission (E>100 GeV). The relatively narrow field of view, few square degrees, of IACTs poses a challenge in the detection of the counterpart in the large uncertainty region provided by the interferometers. A well designed and optimised observational strategy is therefore needed to maximize the probability to select the region hosting the counterpart. Among the IACT experiments, the MAGIC collaboration joined the EM follow-up community in 2014, performing the first follow-up of a GW candidate event (GW151226 [4]) by a Cherenkov Telescope. For the current observing run (O3), MAGIC is optimising the GW follow-up strategy taking into account the information provided with the GW alert, the observational constraints, and taking advantage of its fast repositioning system adopted for Gamma-Ray Bursts (GRBs). In this contribution we will present the MAGIC follow-up strategy of GW events, showing different observational approaches depending on the information available from interferometers and EM instruments.