I consider some of the issues we face in trying to understand dark energy. Huge fluctuations in the unknown dark energy equation of state can be hidden in distance data, so I argue that model-independent tests which signal if the cosmological constant is wrong are valuable. These can be constructed to remove degeneracies with the cosmological parameters. Gravitational effects can play an important role. Even small inhomogeneity clouds our ability to say something definite about dark energy. I discuss how the averaging problem confuses our potential understanding of dark energy by considering the backreaction from density perturbations to second-order in the concordance model: this effect leads to at least a 10\% increase in the dynamical value of the deceleration parameter, and could be significantly higher owing to a UV divergence. Large Hubble-scale inhomogeneity has not been investigated in detail, and could conceivably be the cause of apparent cosmic acceleration. I discuss void models which defy the Copernican principle in our Hubble patch can explain acceleration through inhomogeneous cosmic curvature. These can fit the small scale CMB, and can explain the observed primordial lithium abundances - a niggling 4 or 5 sigma discrepancy in the concordance model. I describe how we can potentially rule out these models, and so provide an important test for the existence of dark energy.