Groundwater regimes are key determinants influencing the outcome of peatland ecosystem restoration programmes. Despite widespread application, impacts of ditch blocking on water table levels (WTLs) remain poorly quantified. Near-continuous WTL monitoring at 21 locations on a degraded Atlantic blanket bog, identified using distributed hydrological modelling, for 21 months prior to ditch blocking, and 24 months post-blockage, aimed to determine how restoration works impacted groundwater levels. Data analysis, using machine learning (ML), trained on baseline WTL and meteorological data, revealed significant water level rises, attributable to ditch blocking. Median summer water levels proved shallower (≥1 cm) at 89% of monitored locations, with an average median summer WTL increase of 4 cm (± 6 cm). Moreover, post-restoration responses to rainfall events indicated slower WTL recession during subsequent dry periods. Findings suggested drain orientation, surface slope and contributing catchment area significantly influenced WTL response. The most significant post-restoration changes in WTL occurred on slopes with gradients <3%. Blocking ditches running perpendicular to slopes >3% had more limited impact, unless they had a large contributing catchment area. The approach adopted demonstrated the potential of applying ML with distributive modelling to peatland restoration programmes, by providing quantifying WTL responses to restoration works under variable meteorological conditions. More generally, study findings pointed to the importance appropriate siting of monitoring points, collection of baseline data and relevant hyperparameter selection for more confidently interpreting post-restoration impacts.  The approach adopted provides a means of managing restoration works to generate the greater hydrological impact in programmes where resources are limited.