Trismus is caused by damage to the masticatory muscles resulting from cancer or its treatment. Contouring these muscles allows avoidance planning to reduce radiation related toxicity. Clinician contouring is however undermined by inter-observer variability. A validated contouring atlas for the masticatory muscles would help standardise volumes, reduce inter-observer variability and improve accuracy. This study aimed to evaluate a novel in-house contouring atlas for the masticatory muscles and assess its ability to reduce inter-observer variability. An atlas was developed by a multi-disciplinary team including a Consultant Radiologist, a Maxillo-facial surgeon and Clinical Oncologists. The muscles of mastication in the atlas were: medial and lateral pterygoids (MP, LP), masseter (M), temporalis (T) as well as the temporo-mandibular joint (TMJ). Seven head and neck clinicians delineated five paired muscles of mastication on CT scans from 5 patients without the atlas. Structures were delineated in a treatment planning system. After a minimum gap of four weeks clinicians were provided with the atlas and asked to re-contour the structures. In-house software was used to analyse the variation between clinicians. Using contours generated by clinical expertise on the same 5 CT scans as the reference, the dice similarity coefficient (DSC), mean distance-to-agreement (DTA) and the distance between the centre of mass (COM) were calculated and compared for each volume with and without the atlas. The mean values over all patients and the standard deviation (SD) were measured to assess inter-observer variation. Using the atlas significantly improved mean DSC for the right and left MP (0.66 v 0.70 p=0.03; 0.76 v 0.80 p<0.01), T (0.63 v 0.75 p<0.01; 0.68 v 0.78 p=0.02) and TMJ (0.62 v 0.79 p=0.01; 0.64 v 0.78 p<0.01). Mean DTA improved for most muscles but were unchanged for the bilateral LP and M. The largest improvement in mean DTA was seen for the right and left T (1.5 v 5.1 mm, p <0.01; 1.6 v 5.8 mm, p=0.02). The SD of all metrics reduced with the atlas suggesting a reduction in variation between observers. The largest significant reduction in SD for DTA was seen for the right and left T (1.3 v 4.6 mm, p=0.05; 0.7 v 4.6 mm, p=0.03) and the right and left TMJ (0.5 v 1.0 mm, p=0.01; 0.5 v 1.1 mm, p=0.01). Distance between the COM reduced in all directions for the right MP, bilateral T and TMJ with the largest significant improvement seen for the left T in the superior-inferior direction (2.2 v 6.4 mm, p =0.03). A new atlas for contouring masticatory muscles during radiotherapy planning for head and neck cancer reduces inter-observer variability with the largest benefit observed for the Temporalis muscle. Reductions in inter-observer variability and improved identification of structures should improve the accuracy of radiotherapy avoidance planning with the aim of reducing radiation related toxicity.