Purpose <p>To evaluate the accuracy and precision of a web-based smartphone app designed to measure sphero-cylindrical over-refraction.</p> Methods <p>A total of 307 healthy young subjects (22.5 ± 3.4 years) underwent clinical subjective refraction. In addition to each subject's refractive correction, spherical and cylindrical errors from −1.50 to 0.00 D were randomly induced so that their sum did not exceed −2.50 D, thereby generating a clinical over-refraction. Seven different IOS and Android-based smartphones were used. A new app showing tripole rescaling stimuli in blue or red light was used. Each subject used one smartphone to measure their induced over-refraction subjectively, obtaining an app over-refraction finding. Differences between the app and clinical over-refraction were analysed across 675 values.</p> Results <p>The mean difference between the app and clinical over-refraction in terms of power vector components ± limits of agreement was 0.17 ± 0.84 D for M (<i>p</i> &lt; 0.0001), 0.01 ± 0.42 D (<i>p</i> = 0.28) for J₀ and −0.01 ± 0.32 D (<i>p</i> = 0.33) for J₄₅. The corresponding values for sphere and cylinder were 0.16 ± 0.80 D (<i>p</i> &lt; 0.0001) and 0.01 ± 0.85 D (<i>p</i> = 0.40), respectively. The sphere and cylinder measurements obtained by the app differed from the expected values by no more than 0.50 D in 85% of cases.</p> Conclusions <p>The accuracy and precision of the over-refraction obtained by the app was slightly lower than standard clinical subjective refraction in terms of inter-examiner repeatability and comparable to measurements obtained using objective methods. The app shows the potential to perform sphero-cylindrical over-refractions reliably.</p>