Solar geoengineering (SGE) can combat climate change by directly reducing temperatures. Both SGE and the climate itself are surrounded by great uncertainties. Implementing SGE affects learning about these uncertainties. We model endogenous learning over two uncertainties: the sensitivity of temperatures to carbon concentrations (the climate sensitivity), and the effectiveness of SGE in lowering temperatures. We present both theoretical and simulation results from an integrated assessment model, focusing on the informational value of SGE experimentation. Surprisingly, under current calibrated conditions, SGE deployment slows learning, causing a less informed decision. For any reasonably sized experimental SGE deployment, the temperature change becomes closer to zero, and thus more obscured by noisy weather shocks. Still, some SGE use is optimal despite, not because of, its informational value. The optimal amount of SGE is very sensitive to beliefs about both uncertainties.