The unique properties of green fluorescent protein (GFP) have been harnessed in a variety of bioimaging techniques, revolutionizing many areas of the life sciences. Molecular-level understanding of the underlying photophysics provides an advantage in the design of new fluorescent proteins (FPs) with improved properties; however, because of its complexity, many aspects of the GFP photocycle remain unknown. In this Account, we discuss computational studies of FPs and their chromophores that provide qualitative insights into mechanistic details of their photocycle and the structural basis for their optical properties. In a reductionist framework, studies of well-defined model systems (such as isolated chromophores) help to understand their intrinsic properties, while calculations including protein matrix and/or solvent demonstrate, on the atomic level, how these properties are modulated by the environment. An interesting feature of several anionic FP chromophores in the gas phase is their low electron detachment energy. For example, the bright excited ππ* state of the model GFP chromophore (2.6 eV) lies above the electron detachment continuum (2.5 eV). Thus, the excited state is metastable with respect to electron detachment. This autoionizing character needs to be taken into account in interpreting gas-phase measurements and is very difficult to describe computationally. Solvation (and even microsolvation by a single water molecule) stabilizes the anionic states enough such that the resonance excited state becomes bound. However, even in stabilizing environments (such as protein or solution), the anionic chromophores have relatively low oxidation potentials and can act as light-induced electron donors. Protein appears to affect excitation energies very little (<0.1 eV), but alters ionization or electron detachment energies by several electron volts. Solvents (especially polar ones) have a pronounced effect on the chromophore's electronic states; for example, the absorption wavelength changes considerably, the ground-state barrier for cis-trans isomerization is reduced, and fluorescence quantum yield drops dramatically. Calculations reveal that these effects can be explained in terms of electrostatic interactions and polarization, as well as specific interactions such as hydrogen bonding. The availability of efficient computer implementations of predictive electronic structure methods is essential. Important challenges include developing faster codes (to enable better equilibrium sampling and excited-state dynamics modeling), creating algorithms for properties calculations (such as nonlinear optical properties), extending standard excited-state methods to autoionizing (resonance) states, and developing accurate QM/MM schemes. The results of sophisticated first-principle calculations can be interpreted in terms of simpler, qualitative molecular orbital models to explain general trends. In particular, an essential feature of the anionic GFP chromophore is an almost perfect resonance (mesomeric) interaction between two Lewis structures, giving rise to charge delocalization, bond-order scrambling, and, most importantly, allylic frontier molecular orbitals spanning the methine bridge. We demonstrate that a three-center Hückel-like model provides a useful framework for understanding properties of FPs. It can explain changes in absorption wavelength upon protonation or other structural modifications of the chromophore, the magnitude of transition dipole moment, barriers to isomerization, and even non-Condon effects in one- and two-photon absorption.