Measurements and theoretical calculations of meson form factors are essential for our understanding of internal hadron structure and QCD, the dynamics that bind the quarks in hadrons. The pion electromagnetic form factor has been measured at small space-like momentum transfer $|q^2| < 0.3$~GeV$^2$ by pion scattering from atomic electrons and at values up to $2.5$~GeV$^2$ by scattering electrons from the pion cloud around a proton. On the other hand, in the limit of very large (or infinite) $Q^2=-q^2$, perturbation theory is applicable. This leaves a gap in the intermediate $Q^2$ where the form factors are not known. As a part of their 12 GeV upgrade Jefferson Lab will measure pion and kaon form factors in this intermediate region, up to $Q^2$ of $6$~GeV$^2$. This is then an ideal opportunity for lattice QCD to make an accurate prediction ahead of the experimental results. Lattice QCD provides a from-first-principles approach to calculate form factors, and the challenge here is to control the statistical and systematic uncertainties as errors grow when going to higher $Q^2$ values. Here we report on a calculation that tests the method using an $eta_s$ meson, a heavy pion made of strange quarks, and also present preliminary results for kaon and pion form factors. We use the $n_f=2+1+1$ ensembles made by the MILC collaboration and Highly Improved Staggered Quarks, which allows us to obtain high statistics. The HISQ action is also designed to have small discretisation errors. Using several light quark masses and lattice spacings allows us to control the chiral and continuum extrapolation and keep systematic errors in check.