Variability is an important tool for the diagnostics of physical processes in AGN. The extended lifetime of ISO provides us with a unique opportunity to study the variability of well-defined subsets of sources in order to distinguish between competing non-thermal and thermal emission mechanisms in the infrared. We propose to obtain second epoch photometric data with ISOPHOT for 54 AGNs already observed in cycle 1. We will then compare the variability characteristics of the sub-samples. Thermal emission from dust cannot vary on short time scales, as the emission medium is very extended. On the other hand, non-thermal emission mechanisms like synchrotron emission are able to vary significantly on very short time scales. With two observations (including cycle 1) we expect to find statistical evidence for variability in objects dominated by non-thermal processes. The selected sample contains 27 radio-quiet objects and 27 radio-loud objects. Radio-quiet sources are thought to be dominated by dust emission. Variations on the time-scale of a year are therefore not expected. However synchrotron emission is the usual explanation for the far-infrared and millimetre emission of radio-loud sources, which should therefore exhibit variability. This simple prediction is central to the models and very important to test. This proposal aims at testing this paradigm in a very efficient way. We build a sample from the objects observed in the first cycle by Wilkes et al., Courvoisier et al., and Chini et al. Pooling together these samples allows us to obtain statistically meaningful samples in both radio-quiet and radio-loud categories. We propose to observe the objects in 3 broad-band filters with ISOPHOT from 10 to 200 microns depending on the object to test the variability of the AGN as well as detect any spectral dependence.