Working memory impairment is prevalent in brain injured patients across lesion aetiologies and severities. Unfortunately, rehabilitation efforts for this impairment have hitherto yielded small or no effects. Here we show in a randomized actively controlled trial that working memory performance can be effectively restored by suggesting to hypnotized patients that they have regained their pre-injury level of working memory functioning. Following four 1-h sessions, 27 patients had a medium-sized improvement relative to 22 active controls (Bayes factors of 342 and 37.5 on the two aggregate outcome measures) and a very large improvement relative to 19 passive controls (Bayes factor = 1.7 x 10^13 ). This was a long-term effect as revealed by no deterioration following a 6.7 week no-contact period (Bayes factors = 7.1 and 1.3 in favour of no change). To control for participant-specific effects, the active control group was crossed over to the working memory suggestion and showed superior improvement. By the end of the study, both groups reached a performance level at or above the healthy population mean with standardized mean differences between 1.55 and 2.03 relative to the passive control group. We conclude that, if framed correctly, hypnotic suggestion can effectively improve working memory following acquired brain injury. The speed and consistency with which this improvement occurred, indicate that there may be a residual capacity for normal information processing in the injured brain.