Estuarine salt intrusion causes problems with freshwater availability in many deltas. Water managers require timely and accurate forecasts to be able to mitigate and adapt to salt intrusion. Data-driven models derived with machine learning are ideally suited for this, as they can mimic complex non-linear systems and are computationally efficient. We set up a long short-term memory (LSTM) model to forecast salt intrusion in the Rhine-Meuse delta, the Netherlands. Inputs for this model are chloride concentrations, water levels, discharges and wind speed, measured at nine locations. It forecasts daily minimum, mean and maximum chloride concentrations up to 7 d ahead at Krimpen aan den IJssel, an important location for freshwater provision. The model forecasts baseline concentrations and peak timing well but peak height is underestimated, a problem that becomes worse with increasing lead time. Between lead times of 1 and 7 d, forecast precision declines from 0.9 to 0.7 and forecast recall declines from 0.7 to 0.5 on average. Given these results, we aim to extend the model to other locations in the delta. We expect that a similar setup can work in other deltas, especially those with a similar or simpler channel network.