• Make your assessments authentic, personalised and meaningful to the students. Arnold (2020) defines authentic assessment activities as characterised by realism, cognitive challenge and evaluative judgement with relevance to self, discipline or professional community. 
• Use alternatives to essays e.g. presentations or discussions/role plays. Presentations offer the chance to question students about the content. 
• Require evidence of the process – regular updates/blog posts or video diary. 
• Include reflective activities asking students to reflect on their own learning, giving specific contexts that AI won’t be able to reference. 
• Scaffold the assessment – building it up over time. 
• Avoid re-using assessments. 
• Avoid generic essay titles that are easily searched with a search engine – be really specific and set a narrow context. 
• Use the standard AI Use Level Descriptors so that students are clear about what is acceptable.
• Try using AI to answer your assessment brief as you develop it. If it produces something reasonable, you may need to come up with something else.

Updates to this advice

Previously, we suggested making assessments topical because the AI models are trained with data which is a few years out of date and would not know about recent events. OpenAI have announced updates to ChatGPT which mean that this advice no longer applies.