Next-generation neurostimulators capable of running closed-loop adaptive deep brain stimulation (aDBS) are about to enter the clinical landscape for the treatment of Parkinson's disease. Already promising results using aDBS have been achieved for symptoms such as bradykinesia, rigidity and motor fluctuations. However, the heterogeneity of freezing of gait (FoG) with its wide range of clinical presentations and its exacerbation with cognitive and emotional load make it more difficult to predict and treat. Currently, a successful aDBS strategy to ameliorate FoG lacks a robust oscillatory biomarker. Furthermore, the technical implementation of suppressing an upcoming FoG episode in real-time represents a significant technical challenge. This review describes the neurophysiological signals underpinning FoG and explains how aDBS is currently being implemented. Furthermore, we offer a discussion addressing both theoretical and practical areas that will need to be resolved if we are going to be able to unlock the full potential of aDBS to treat FoG.