In a randomised controlled trial, we showed that a five-day intervention combining anodal transcranial direct current stimulation over the left inferior frontal cortex with temporary speech fluency enhancing techniques reduces stuttering. Speech fluency was unchanged by the fluency training alone, as predicted. Here, we report the neural changes associated with the intervention, measured using functional MRI during sentence reading before the training and one-week later. We obtained imaging data in 25 adult men who stutter (median age = 32 y, inter-quartile range = 11) at the pre-intervention baseline and again one-week post-intervention. A control group of 15 adult men who do not stutter (median age = 30 y, inter-quartile range = 10) and did not complete the intervention were scanned on one occasion. In a whole-brain analysis of perceptibly fluent sentence reading, we compared the change in task-evoked neural activity in the sub-group of men who stutter who had received active stimulation during the intervention (N=13) with those who had sham stimulation (N=12). We hypothesised that the combination of anodal stimulation over the left inferior frontal cortex and fluency-enhancing training would result in lasting change to the brain networks supporting fluent speech production. An additional region-of-interest analysis explored effects on basal ganglia nuclei, which are thought to have a key role in the casual mechanism of stuttering, and which we hypothesised would be engaged by the behavioural approach used during training (choral and metronome-timed speaking). One week after the intervention, the group who had received active transcranial stimulation showed increased activity in speech-related brain regions, relative to the group who had received sham stimulation. Cortically, these changes were evident in left inferior frontal cortex (pars opercularis and orbitalis), anterior insula, anterior superior temporal gyrus, anterior cingulate cortex, and supplementary motor area. Subcortically, activation increased in the caudate nuclei and putamen bilaterally, and in right globus pallidus and thalamus. Together these regions form cortico-striatal-thalamo-cortical loops involved in the planning and initiation and control of speech movements.Our findings reveal that the mechanism of action of the tDCS intervention involved increasing activity across the network involved in the production of fluent speech, indicating that tDCS can be used to promote neural plasticity to strengthen networks supporting natural fluency. This study advances the potential of using non-invasive brain stimulation to improve therapy efficacy for those people who stutter who choose to work on increasing fluency.