This research focused on refining alkali activator parameters to enhance the mechanical performance and microstructural characteristics of diatomaceous earth (DE) based geopolymer paste. The influence of sodium hydroxide (NaOH) concentration (10, 12, and 14 M) and the sodium silicate to sodium hydroxide ratio (Na₂SiO₃/NaOH) (0.5, 1.0, and 2.0) on key response variables was evaluated using Response Surface Methodology (RSM), including compressive strength, flexural strength, flowability, and initial setting time. The optimal mixture was identified at a NaOH concentration of 10 M and a sodium silicate proportion of 53.04% in the total activator solution, resulting in a compressive strength of 18.05 MPa and a flexural strength of 2.32 MPa. Microstructural investigations using SEM–EDS, XRD, and FTIR confirmed the formation of a compact geopolymer matrix with low porosity, predominantly composed of sodium aluminosilicate hydrate (N–A–S–H) gel. ANOVA results indicated that the Na₂SiO₃ to total activator ratio had a more significant effect on mechanical properties (p < 0.05) than NaOH molarity. Furthermore, the empirical model demonstrated strong predictive reliability, with deviations between predicted and experimental values remaining below 5%. The minimal errors in flow (0.74%) and initial setting time (1.51%) reflect accurate prediction of fresh state behavior and reaction kinetics, while the low errors in compressive (2.95%) and flexural strength (4.04%) confirm the model’s capability to estimate hardened state performance. Overall, the optimized alkali activation regime significantly improved the geopolymerization of DE, supporting its potential as a sustainable, locally sourced construction material with enhanced performance. This study presents a novel approach by simultaneously optimizing NaOH molarity and Na₂SiO₃/NaOH ratio using response surface methodology (RSM) to enhance both mechanical performance and microstructural characteristics of diatomaceous earth–based geopolymer paste, providing a comprehensive optimization framework not previously reported.