Background/Aims Peak hip adduction and knee flexion during running are associated with patellofemoral pain persistence, representing treatment targets. Clinical practice is lacking a validated, reliable tool with which to measure these kinematics. This study aimed to determine the accuracy of clinical gait analysis, by investigating concurrent validity, intra- and inter-rater reliability of two-dimensional video. Methods A total of 21 participants with patellofemoral pain were recruited (10 males, 11 females). Synchronised three-dimensional and two-dimensional kinematic data were collected during over-groundrunning. Two-dimensional videos were analysed with the Hudl Technique application using a commercially available tablet (iPad). Single measure intraclass coreelation coefficients (ICCs) were calculated using a two-way mixed effects model with absolute agreement. Three-dimensional peak hip internal rotation was investigated as a covariate with backward linear regression, using the F change statistic. Results There was poor agreement between three-dimensional and two-dimensional measurement of peak hip adduction (ICC 0.06) and peak knee flexion (ICC 0.42). Moderate intra-rater reliability was identified for both variables (ICC 0.61–0.65). Inter-rater reliability for peak knee flexion was moderate (ICC 0.71), but was poor for peak hip adduction (ICC 0.31). Three-dimensional peak hip internal rotation did not significantly explain the identified poor agreement for either variable. Conclusions Poor agreement between three-dimensional kinematics and two-dimensional video was identified for both variables in runners with patellofemoral pain, despite acceptable intra-rater reliability. Investigation of software with increased precision is warranted, to improve the accuracy of two-dimensional video predicting three-dimensional kinematics in the clinical setting. Clinical gait analysis using the Hudl Technique application is not currently advocated.