The interest of geophysical prospection methods for the study of the first Iron Age settlements is no longer in question. Numerous flagship sites of the Celtic period have benefited from this type of analysis and confirm this strong interest. If these studies are often illustrated by marvelous pictures, they tend to minimize the difficulties encountered by geophysics: types of archaeological features, but also on certain sites for which inconclusive results are often sidelined. These difficulties are heightened in the singular context of preventive archaeology. Here, exceptional sites rub shoulders with more modest or less well-preserved settlements, where one must try to understand, in a short time, the labile traces observed after stripping. When they are subject to geophysical mapping before stripping, these same sites highlight the limits of these methods and require geophysical analysis to be pushed to its limits. The intrinsic uncertainty of geophysical interpretations must then face the reality of field observations. Although certain geophysical anomalies enrich or complete these observations, geophysics can also prove to be inefficient on features that are obvious during excavation, whether by their dimensions, their state of conservation or the richness of the material they are likely to contain. It is then necessary to weigh the hypotheses and the proposed geophysical interpretations while drawing a maximum of information from the acquired data and by developing new applications such as the use of geophysics on stripped areas. We will present here, through several case studies, some methodological limits, interpretation mistakes, but also the strong potential of these methods for the understanding of both spatial organizations on large or small scales, and functional organizations especially illustrated by the analysis on stripped areas.