Optogenetics is one of the most important techniques in neurophysiology, with potential clinical applications. However, the strong light needed may cause harmful temperature rises. So far, there are no methods to reliably estimate brain heating and safe limits in actual optogenetic experiments. We used thermal imaging to directly measure such temperature rises at the surface of live mouse brains during laser illumination with wavelengths and intensities typical for optogenetics. We then modeled the temperature rise with a simple logarithmic model. Our results indicate that previous finite-element models can underestimate temperature increases by an order of magnitude. We validate our empirical model by predicting the temperature rise caused by pulsed stimulation paradigms. These predictions fit closely to the empirical data and constitute a better estimate of real temperature increases. Additionally, we provide a web-based app for easy calculation that can be used as a tool for safe design of optogenetic experiments.