The expected fraction of astrometric planet candidates that are genuine planets (instead of unresolved binaries) as a function of apparent stellar mass M_⋆, G-band magnitude, and semimajor axis a. For most values of M_⋆, G, and a, our simulations predict that ≳80% of candidates will be genuine planets. The expected fraction of candidates that are unresolved binaries is largest for low-mass stars, faint stars, and small orbital separations, in which case it can exceed 50%.