Organophosphate chemicals such as herbicides, pesticides, and nerve agents are harmful even in small amounts, thus requiring methods for rapid ultrasensitive detection. Surface-enhanced Raman spectroscopy (SERS) is a leading method for trace-level detection of dissolved chemicals, yet our understanding of the critical influence of the solvent is currently incomplete. Rather than being recognized as a crucial aspect for generating a strong signal enhancement, the solvent is often dismissed as a secondary parameter. This has left the dual impact of solvents on the analyte and SERS substrate properties in relation to SERS largely unexplored. In this work, we evaluated the suitability of common solvents for the detection of nerve agents using SERS on gold nanopillar SERS substrates. Overall, we found large differences in the limit of detection (LOD) between solvents. VX exhibited an LOD of 0.2 ppm in water, compared to 8.7 and 6.3 ppm in ethanol and acetonitrile, respectively, with similar trends for other nerve agents. Dichloromethane and n-hexane were found to be largely ineffective as solvents for SERS in these applications. To explain these differences, we used DFT to model analyte-solvent interactions, showing that high-polarity solvents such as water generate higher electron concentrations in select analyte molecular groups, impacting analyte-substrate interactions. Furthermore, we used Raman mapping, contact angle measurements, and SEM to evaluate solvent-substrate interactions, showing that solvent-dependent capillary forces impact SERS substrate nanogeometry and thus the formation of SERS hotspots. Overall, these results provide insight into the many solvent effects that are relevant to designing future SERS sensing assays.