This dataset comprises EEG recordings from 8 ALS patients performing a P300-based brain-computer interface speller task using a 9x8 character grid. Participants completed 2 sessions each, with 16-channel EEG data acquired at 256 Hz using dry and wet electrode configurations for comparison. The dataset is part of the BigP3BCI study, the largest public P300 BCI dataset, and includes standardized HED event annotations for target and non-target visual stimuli.