We analysed six Apertif datasets, covering the GOODS-N LOFAR deep field region, aiming to improve our understanding of the faint radio source composition, their polarisation behaviour, and how this affects our interpretation of polarised source counts. Using a semi-automatic routine, we ran rotation measure synthesis to generate a polarised intensity mosaic for each observation. The routine also performs source finding and cross-matching with the total power catalogue, as well as NVSS, SDSS and allWISE, to obtain a catalogue of 1182 polarised sources in an area of 47.4deg^2^. Using the mid-infrared (MIR) radio correlation, we found no indication of any polarised emission from star formation. To robustly estimate the source counts, we performed an investigation of our sample's completeness as a function of the polarised flux via synthetic source injection. In contrast to previous works, we find no strong dependency of fractional polarisation on the total intensity flux density. We argue that differences regarding previous (small-scale, deep field) analyses can be attributed to sample variance. Relative to the findings of previous works, here we find a steeper slope for our Euclidean-normalised differential source counts. This is also visible as a flattening in cumulative source counts. We attribute the observed steeper slope in Euclidean normalised differential source counts to a change in the source composition and properties at low total intensities.