ABSTRACT We present multi-epoch spectroscopic follow-up of a sample of ellipsoidal variables selected from Gaia Data Release 3 (DR3) as candidates for hosting quiescent black holes (BHs) and neutron stars (NSs). Our targets were identified as BH/NS candidates because their optical light curves – when interpreted with models that attribute variability to tidal distortion of a star by a companion that contributes negligible light – suggest that the companions are compact objects. From the likely BH/NS candidates identified in recent work accompanying Gaia DR3, we select 14 of the most promising targets for follow-up. We obtained spectra for each object at 2–10 epochs, strategically observing near conjunction to best constrain the radial velocity semi-amplitude. From the measured semi-amplitudes of the radial velocity curves, we derive minimum companion masses of $M_{2,\, \rm min} \le 0.5 \, {\rm M}_{\odot }$ in all cases. Assuming random inclinations, the typical inferred companion mass is $M_2 \sim 0.15\, {\rm M}_{\odot }$. This makes it unlikely that any of these systems contain a BH or NS, and we consider alternative explanations for the observed variability. We can best reproduce the observed light curves and radial velocities with models for unequal-mass contact binaries with star-spots. Some of the objects in our sample may also be detached main-sequence binaries, or even single stars with pulsations or star-spot variability masquerading as ellipsoidal variation. We provide recommendations for future spectroscopic efforts to further characterize this sample and more generally to search for compact object companions in close binaries.