Hybrid organic–inorganic perovskites (HOIPs), such as CH₃NH₃PbI₃ (MAPbI₃), are attractive for inexpensive, high-performance solar cells. Controlling HOIP thin-film quality and morphology, which is essential to achieve consistent solar-cell efficiencies, requires a fundamental understanding of the link between solution chemistry and crystallization pathways. To elucidate the effect of solvent choice and solution speciation on the crystallization pathway, we combined computational modeling of molecular-level solvent–solute interactions in HOIP growth solutions with experimental monitoring of film phase evolution. Using density functional theory calculations and a Bayesian optimization-based approach (PAL), we exhaustively searched the HOIP/solvent combinatorial space to obtain a ranked list of increasing HOIP/solvent intermolecular binding energy. Then, using in situ X-ray diffraction, we tested solvents of varying coordinating abilities with MAPbI₃ to correlate the PAL-generated ranking with the crystallization pathway. Weakly coordinating solvents (e.g., N-methyl-2-pyrrolidone) formed the perovskite via a two-step crystallization pathway, where a crystalline intermediate formed and was subsequently transformed to perovskite with heating, whereas strongly coordinating solvents (e.g., tetrahydrothiophene 1-oxide) formed the perovskite directly from solution. We propose that the solution coordination chemistry determines the crystallization pathway. Our integrated experimental–computational approach could be applied to study the interplay between solution chemistry and crystallization pathways for other solution-grown materials.