This article develops a machine learning methodology to analyse the relationship between university accreditation and student experience. It is applied to 98 university accreditations conducted by the Quality Assurance Agency (QAA) in the UK in 2012-2018, and 263,025 university ratings in three categories posted by students on the website whatuni.com. Natural Language Processing (NLP) is used to extract features from the accreditation reports. These features are explanatory variables in automated linear regression models where the dependent variable is the student experience, as measured by the student ratings. It finds that the Institutional Reviews in 2012-2014 and the Higher Education Reviews in 2014-2016 misinform the public about the student experience, while the Enhancement-Led Institutional Reviews in Scotland in 2014-2018 provide sound guidance. These findings should lead to a deep reflection on how the university accreditation system functions in the UK. They also contribute to the ongoing debates on student engagement in HE quality assurance and whether student experience is a reliable measure of university quality. Finally, it is shown that machine learning models are useful tools to compare accreditation reports and can assist prospective students in choosing the university.