In Malaysian universities, many educators are spending hours grading essays and theses suspected of being largely generated by artificial intelligence (AI), raising the question of who is actually being assessed. The article notes that getting an A- or even a B+ has become the new abnormality, often followed by student emails asking why they fell short of a perfect A. Traditionally, Bloom's Taxonomy has guided learning, where students progress from remembering vocabulary and grammar rules to creating an original essay, which was seen as evidence of deep learning. However, today, "creation" can be reached with a simple prompt entered into a large language model (LLM), producing a polished essay within seconds, making the summit the baseline. Educators are caught in a game of cat and mouse, relying on AI detectors whose reliability remains questionable. The article argues that if a large portion of the cognitive work can be done by an algorithm, an essay no longer necessarily equates to deep learning. A proposed solution is a reversed Bloom’s Taxonomy, popularised by educator Michelle Kassorla, which suggests that real proof of learning begins after the AI generates the first draft. This would involve asking students to identify factual inaccuracies or hallucinations, critique weak reasoning, challenge unsupported claims, and verbally defend their revisions. The author argues that universities must redesign assessments to measure students’ understanding, critical thinking, and ability to defend their work, rather than simply the polished outputs that AI can generate.