The BGS Predictive Seabed Sediments UK v1 dataset includes digital maps of seabed sediment (SBS) composition across the UK Continental Shelf (UKCS). The dataset includes one classified SBS map (Folk sediment class; vector), and three percentage-sediment maps (rasters) giving the proportions of gravel, sand, and mud. The dataset is generated using a machine learning algorithm known as a Distributional Random Forest (DRF). The model uses input data to predictively classify what seabed sediments are most likely to be the dominant sediment present. It does so, using more than 38,000 seabed sediment samples (collated from various sources) with measurements of the proportion of mud, sand and gravel from locations across the study area. The predictions are constrained against covariate variables that are known to influence which sediment is most likely to occur. These include bathymetry data morphometric derivatives (at multiple spatial scales), as well as hydrodynamic data layers (currents and tidal). The dataset was reviewed via a qualitative assessment (QA) protocol by subject-area experts (e.g. contrasting with previous mapping, and local examples of higher-resolution data and mapping), and following methodological improvements based on this feedback, updated SBS map products were prepared. The dataset is presented at a national-scale, with a spatial resolution of approximately 110m, covering the UKCS (slightly modified UKCS area based on data availability).