We search for gravitational wave (GW) events from LIGO-Virgo's third run that may have been affected by gravitational lensing. Gravitational lensing delays the arrival of GWs, and alters their amplitude -- thus biasing the inferred progenitor masses. This would provide a physically well-understood interpretation of GW detections in the "mass gap" between neutron stars and black holes, as gravitationally lensed binary neutron star (BNS) mergers. We selected three GW detections in LIGO-Virgo's third run for which the probability of at least one of the constituent compact objects being in the mass gap was reported as high with low latency -- i.e. candidate lensed BNS mergers. Our observations of powerful strong lensing clusters located adjacent to the peak of their sky localisation error maps reached a sensitivity $\rm AB\simeq25.5$ in the $z'$-band with the GMOS instruments on the Gemini telescopes, and detected no candidate lensed optical counterparts. We combine recent kilonova lightcurve models with recent predictions of the lensed BNS population and the properties of the objects that we followed up to show that realistic optical counterparts were detectable in our observations. Further detailed analysis of two of the candidates suggests that they are a plausible pair of images of the same low-mass binary black hole merger, lensed by a local galaxy or small group of galaxies. This further underlines that access to accurate mass information with low latency would improve the efficiency of candidate lensed NS-NS selection.