Marine coastal ecosystems are facing structural and functional changes due to the increasing human footprint worldwide, and the assessment of their long-term changes becomes particularly challenging. Measures of change can be done by comparing the observed ecosystem status to a purposely defined reference condition. In this paper, a geospatial modelling approach based on 2D mapping and morphodynamic data was used to predict the natural position of the upper limit (i.e., the landward continuous front) of Posidonia oceanica seagrass meadows settled on soft bottom. This predictive model, formerly developed at the regional spatial scale, was here applied for the first time at the Mediterranean spatial scale in eight coastal areas of Spain, France, Italy, and Greece showing different coastal morphologies and hydrodynamic characteristics, and affected by a number of natural and/or human local disturbances. The model was effective in measuring the regression (i.e., seaward withdrawal) of the meadow upper limit. In all the meadows investigated the upper limit was regressed, laying deeper than the reference condition, with the proportion of regression ranging from 17.7% to 98.9%. The highest values of regression were found in Spain and in France, and were consistent with the highest levels of fragmentation detected with map analysis and of coastal pressures. This geospatial modelling approach represents an effective tool to define the reference conditions when proper pristine areas or historical data are not available, thus allowing the assessment of long-time changes experienced by seagrass ecosystems due to human impacts.