It is widely appreciated that cold, dense molecular gas is very difficult to detect. Modelling has demonstrated that such gas can exist in the form of spherical, self-gravitating clouds with masses in the planetary range and radii in the AU range. Microlensing studies — both Galactic and extragalactic — are one of the main ways in which such objects can be revealed, or constrained, but the task is complicated by the non-point-like nature of the lenses. I will describe the microlensing signatures that are predicted by our physical models of cloud structure — including the effects of gravitational lensing, gas lensing and extinction — and the constraints that are imposed by the available data.