Cognitive impairment is a common consequence of stroke, but remains difficult to predict. We investigate the ability of early QEEG assessment to inform such prediction, using binary logistic regression. Thirty‐five patients (12 female, ages 18–87) suffering middle cerebral artery, ischemic stroke were studied. Resting‐state EEG was recorded 48–239 h after symptom onset. Relative power for delta, theta, alpha, and beta bands, delta:alpha ratio, and peak alpha frequency were analyzed. Montreal Cognitive Assessment (MoCA) was administered, where possible, on day of EEG and at median 99 days (range 69–138) poststroke. Eight patients could not complete the baseline MoCA, and four the follow‐up MoCA, for varying reasons (most commonly, stroke symptoms). Fifteen patients (48%) had cognitive impairment (MoCA score ≤25) at follow‐up. One QEEG index was able to correctly predict presence/absence of cognitive impairment in 24/31 patients (77.4%), whereas predischarge MoCA did so in 23 patients. This index, relative theta frequency (4–7.5 Hz) power, was computed from only three posterior electrodes over the stroke‐affected hemisphere. Its predictive accuracy (three electrodes) was higher than that of any “global” QEEG measure (averaged over 19 electrodes). These results may signify association between poststroke alpha slowing and cognitive impairment, which may be mediated by attentional (dys)function, which warrants further investigation. Pending further studies, QEEG measure(s)—from a few electrodes—could inform early prognostication of poststroke cognitive outcomes (and clinical decisions), particularly when cognitive function cannot be adequately assessed (due to symptoms, language, or other issues) or when assessment is equivocal.