In this work we address a problem of active fault detection in an IoT scenario, whereby a monitor probes a remote device in order to detect faults and acquire fresh information. However, probing can have a significant impact on the IoT networks energy and communication resources. To address this problem we utilize Age of Information as a measure of the freshness of information at the monitor and adopt a semanticsaware communication approach between the monitor and the remote device. In semantics-aware communications, the processes of generating and transmitting information are treated jointly in order to consider the importance of information and the purpose of communication. We formulate the problem as a Partially Observable Markov Decision Process and propose a computationally efficient stochastic approximation algorithm to approximate the optimal policy. Finally, we present numerical results that exhibit the advantage of our approach compared to a conventional delay-based probing policy.