Observing active nuclei, Ultra-Luminous Infrared Galaxies, starburst and merging galaxies, is both a challenge and a requirement for adaptive optics. It is a requirement, because models needed to explain the high infrared flux and the physics of these monsters need constraints that come, in part, from the fine details gleaned on high angular resolution images, and it is a challenge because, being distant, these objects are usually faint in apparent visual magnitude, meaning that the wavefront sensors have to operate in a photon starved regime. Many observations have been controversial in the past, and it is always difficult to tell an artifact such as astigmatism from an inner bar. The importance of observing the point spread function is therefore even more crucial than on bright objects, as PSF reconstruction methods 'a la Veran' break down when the photon noise dominates the statistics of the wave front, or when locking the loop on extended objects. Yet, while some cases have been controversial, some very clear and profound results have been obtained in the extragalactic domain, such as the detection of host galaxy to quasars and star formation studies. It turns out that the fundamental prerequisite to such success stories is a stable, well understood and well calibrated PSF.