The Ubaye Region is a seismically active region in the Western Alps (France), regularly struck by seismic swarms characterized by a high number of small to moderate earthquakes, such as in 2003–2004 or 2012–2015. While some earthquakes could be associated with known faults, the character of the observations (high seismicity – low deformation rate) requires complex driving processes beyond local or regional tectonics. Most conceptual models involve fluids present down to depths of several km, and/or long-range transport. Magnetotellurics (MT) is known to be an efficient imaging method sensitive to crustal fluids. During 2020/21, a data set of 30 MT sites was acquired, covering a signal period ranging between 10-4 to 104 s, with generally all 5 components measured. Data quality was generally satisfactory up to 3 s and sometimes up to 100 s. Major problems were related to topography (including logistics), and to the presence of electromagnetic noise, only to be mitigated by advanced processing methods (FFMT). For the 3-D inversion required by the data (phase tensors, WAL, topography), we have chosen a joint inversion of induction vectors, phase tensors and off-diagonal impedances (previously corrected for static shift with help of phase tensor inversion). This allowed us to obtain the best 3-D model using the ModEM inversion code, explaining all three data types reasonably well. The main findings from this investigation include (a) a prominent conductor (down to 20 Ωm) located along the axis of the swarm zone, though generally above it; (b) a regional dominance of the Penninic Front in the East and the overridden Mesozoic (Dauphinoise) sediments in the West, both not fully covered by the current survey; (c) strike directions that agree well with most of the mapped faults and focal mechanisms of the strongest seismic events. Uncertainties associated with the insufficient data coverage in some of the most interesting zones were studied by analysing the sensitivities provided by the inversion and direct forward modelling of significant model features. In general, this led to the conclusion that our sensitivity does reach the border of the seismic swarm activity, but does not cover its depth extent. Due to the gap in data in the central area of interest, the geometry and connectivity of the main conductor remains a hypothesis. Thus, a truly quantitative characterization of prominent identified structures is not currently possible and should be better assessed with additional measurement sites. The different conceptual models proposed for the origin of the seismic swarm activity will be discussed in the light of the MT imaging, and the associated uncertainties.