Written assignments for large classes pose a far more significant challenge in the age of the GenAI revolution. Suggestions such as oral exams and formative assessments are not always feasible with many students in a class. Therefore, we conducted a study in South Africa and involved 280 Honors students to explore the usefulness of Turnitin's AI detector in conjunction with student self-reflection. Using a Mixed Methods Research (MMR) approach, we analysed data generated from the Turnitin AI reports, our grading rubrics, and qualitative student self-reflection. The findings show that incorporating self-reflection into assessments supports ethical GenAI use and improves the transparency lecturers need for decision-making. A declaration form allowed the students to be upfront about using Generative Artificial Intelligence tools. We found that students who can reflect on their learning relied less on generated content. However, students with high AI detected scores (> 20%) did not adequately reflect on how the tools supported their learning and could not give credible explanations of use. We contribute to the body of knowledge by providing students and academics with examples of responsibly handling AI-detected scores in large-class settings. We present a guided self-reflection and declaration with an AI detector to support students and help lecturers make decisions when grading. We also present a decision tree that lecturers and graders can use when evaluating AI use in assessments.