READABILITY TESTS CONDUCTED ON THE SOUTH AFRICAN VAT ACT UTILISING GENERATIVE AI AS A PRACTICAL SOLUTION

This study sets out to provide some preliminary findings on the level of readability of the VAT Act in South Africa using well established formulae from the literature. The methodology is readability formulae developed in the education literature and as applied to taxation materials, including legislation, drawing upon the approaches used in overseas studies in countries such as Australia and New Zealand. The study also applies Generative AI (Copilot) to assist with rewriting tax materials to improve their readability. The findings reflect those in prior overseas studies, revealing that the South African VAT Act provisions are very challenging to read and understand, generally requiring the reader to be a university graduate. Material from SARS is found to be more readable than the legislation. The findings suggest that like other jurisdictions, considerable effort is needed to redraft the VAT Act and other material in such a way to improve legislative simplicity so as to make the material more accessible to a wider audience to read and understand. The use of Generative AI is shown to significantly improve readability through suggestions for rewriting VAT legislation although this still requires further review to ensure technical accuracy. This is a first time that the readability of the South African VAT Act has been undertaken, providing a further comparison to the findings of the overseas literature. It provides a basis to encourage legislative drafters to rewrite the VAT Act.