When the Artefact Is No Longer Evidence: Assessment Validity and Academic Integrity in the Age of Generative AI

For most of the history of schooling, producing a piece of work was taken as proof that a learner could do the thinking behind it. Generative AI has broken that assumption. A competent essay, solution or report can now be produced in seconds, and agentic tools can complete whole courses inside a learning platform. Schools and universities have responded mostly by trying to detect AI use, a strategy that research shows is unreliable and biased against non-native English writers. This article draws on a practitioner newsletter series written for school leaders in 2026, read alongside research on detection, assessment validity and high- stakes examinations. It argues that the integrity problem is better understood as a validity problem: the question is not whether a student used AI, but whether an assessment still gives trustworthy evidence of what the student can do. The article sets out the evidence on student use and teacher perceptions, explains why detection fails, describes how single high-stakes examinations have become fragile in 2026, and docu‐ ments assessment designs used in practice, including process portfolios, supervised components, local and personal tasks, oral defence, iterative feedback and source constraints. It proposes an "evidence chain" frame‐ work of five questions that any task should answer, and identifies equity risks in the move to oral and process-based assessment that schools need to manage.