The unsupervised take-home essay is effectively extinct. A 2026 survey by the Higher Education Policy Institute found that 94% of undergraduates in the United Kingdom now use generative AI in assessed work, a figure that has jumped from just 3% two years prior. Turnitin data from Australian tertiary institutions between October 2025 and April 2026 shows a similar trend, with over half of submissions containing AI-assisted writing and one in ten being more than 80% AI-generated. The era of testing unaided student writing in isolated, high-stakes environments has ended, forcing universities to rethink how they verify knowledge.
The Retreat and the Detection Trap
Institutional responses have largely been reactive. The Universities of Melbourne and Toronto are phasing out unsupervised take-home essays, while the University of Cape Town aims to have 40% of assessments invigilated by 2027. At Durham University, a professor of philosophy stepped down as chair of his examiners’ board, citing a marking crisis. This retreat to the exam hall is understandable given the anxiety, but it risks ignoring a more fundamental question: are the skills being measured actually relevant to modern professional life?
A second common response is detection-led enforcement, running submissions through AI detectors to flag violations. This approach is quietly collapsing. Detection tools produce false negatives and, more damagingly, false positives that disproportionately affect non-native English speakers. Several institutions have faced appeals from students wrongly accused based solely on a detector score. Even the University of Sydney, which retains detectors, admits that AI use in unsupervised assessment cannot be reliably detected on its own. They use Turnitin’s indicator only alongside other evidence, never as a standalone basis for integrity cases.
Read Also: University name sparks status debate
Redesigning the Classroom for AI Literacy
The third and most constructive response is redesign. The most influential model is the “two-lane” approach developed by Danny Liu and Adam Bridgeman at the University of Sydney. This framework is now embedded in Sydney’s coursework policy and has been adapted by institutions from Auckland to VU Amsterdam. It is also reflected in the guidance of TEQSA, the Australian tertiary education regulator. The logic is straightforward: secure, supervised assessments verify foundational knowledge, while open, unsupervised assessments assume AI use and test how students interact with it.
Since January 2025, Sydney coordinators cannot prohibit AI in unsecured assessments. The rationale is that an unenforceable prohibition teaches students only that rules are theatre. Instead, the goal is to develop the capability to work with AI critically and transparently. This shift can reduce burden; Brunel University’s programme-level reforms cut summative assessment load by roughly two-thirds. In practice, this means students are no longer judged on whether they used AI, but on the quality of their technological direction. The focus shifts from policing the tool to assessing the human’s ability to direct it.
Five Practices That Build Critical Thinking
These behaviors do not happen spontaneously. They must be designed into the assignment. The first is the AI appendix, where students submit a record of their significant AI interactions, including prompts, responses, and commentary on what they accepted or rejected. This converts invisible use into assessable evidence of thinking. The second is structured error-hunting, where students analyze an AI-generated answer to find invented references or misattributed theories. Catching a confident machine in a fabrication teaches the verification imperative faster than any lecture.
Read Also: Ethical Gemstone Spotlight: Explore the Beauty and Sustainability of Different Gemstones
The third practice is the challenge requirement, which explicitly asks students to document at least two moments where they pushed back on the AI.
Shifting the Focus to Judgment and Transparency
There is an irony in the current panic: the technology that supposedly makes critical thinking obsolete actually makes it indispensable. Large language models invent references and express incorrect information with remarkable confidence. Without careful evaluation, these convincing-sounding answers can mislead uncritical students. Stronger students naturally develop verification habits, comparing AI responses with reliable sources and modifying ideas before using them. That gap in judgment, not access to the tool, is what assessment should now measure. It is the skill that will matter for the rest of these students’ working lives, where the value of leaders rests on judging the quality of information produced by intelligent systems.
This does not mean the essay should disappear. In-person assessment has a place, particularly for foundational knowledge. But abandoning the essay altogether would discard the one form best suited to examining how students think over time. Assignments should require students to analyze, evaluate, and apply ideas to unfamiliar situations while showing their working with the machine. The process record is becoming as important as the final product.