We investigate two natural linear estimates of a signal's continuous time Fourier transform (FT) based on non-uniform samples, where sample value and sample time come in pairs where at least the sampling time is a stochastic process. Such stochastic sampling occurs in several applications, such as queue theory and event based sampling. Analytical expressions of mean and covariance are derived for the FT estimate. Intended use of these are to give uncertainty bounds on transforms and spectral estimates, but they are also useful in signal estimation and system identication.