Generative AI produced images are quickly becoming part of how primary teachers prepare and run lessons, but how do we know if these images are scientifically sound? In this session we share findings from our research showing how generative AI can both reinforce known misconceptions and introduce new ones, particularly when teachers ‘personalise’ content for their pupils, which is a specified use case by the DfE for Generative AI use in education.
During the session you'll see worked examples from common primary science topics. You'll then have the chance to generate, examine and discuss images with other delegates, with prepared images available if you don't have a device.
The session is pitched for a range of confidence levels with AI, including those just starting out, while respecting that many primary colleagues already have significant generative AI expertise. You'll leave with a framework and checklist to evaluate generative AI produced images for misconceptions, plus prompt strategies to improve the science quality of what you (and your pupils) create.
Delegates will be able to: • Recognise common misconceptions reinforced or introduced by generative AI produced primary science images. • Use a clear framework and checklist to appraise generative AI visuals before using them in lessons. • Apply prompting strategies that reduce the risk of misconception-rich outputs. • Confidently discuss generative AI image quality with colleagues and pupils as part of wider digital literacy.
Background: Helen Harden, former chemistry lead on the University of York's BEST project and chemistry curriculum consultant, brings extensive expertise in primary and secondary science misconceptions. Victoria Hedlund, founder of GenEd Labs and contributor to the DfE and Chartered College's Safe and Effective Use of Generative AI in Education guidance, brings critical oversight of generative AI in classroom use. Our shared interest in this area is set out in our article “Looking deeper into AI-generated visual explanations” to be published in the July 2026 edition of SSR with a primary-focused article planned for ASE’s Primary Science journal. This forms the precursor to a larger ongoing research project.
