Why Steve Hanke Calls AI Cheating 'Blatant' to Spot

The spread of ChatGPT and similar chatbots has made it easier for students to generate homework, but Johns Hopkins University professor Steve Hanke says he does not rely on detection software to identify machine-written work. Having taught applied economics at the university for nearly 60 years, Hanke called himself an 'old fox' and said he finds it very easy to separate a student's writing from a chatbot's output.

His reasoning is personal and experiential rather than technical. Hanke, who served on President Reagan's Council of Economic Advisers, said many of today's students have weak writing skills, even at elite universities, and that he has a strong sense of each student's economics knowledge. When a submission is out of line with either baseline, he said, it is blatant that the work was produced by AI. The main change, he added, is that AI requires him to be more vigilant than before.

Drusilla Blackman, a former dean of admissions at Harvard and Columbia, offered a similar view in an interview with Business Insider. Blackman, founder of Deans of Admissions, said it is almost always evident to an educator when a student has used AI because the submitted work does not match the student's standard of writing or critical thinking. She said work that is not proportional to that standard is detectable. Separately, teachers previously told Business Insider they are using AI defensively by creating AI-resistant assignments and returning to handwritten work.

What Hanke's Baseline Method Can and Cannot Prove

Baseline knowledge, not AI detection software

Hanke's claim rests on what he says he already has: familiarity with each student's economics ability and written expression. That makes the signal a mismatch with a known baseline, not a technical property of AI-generated text. The article does not provide data on how often this approach produces false positives or misses AI use, so the confidence expressed is based on personal experience rather than systematic evidence.

What Drusilla Blackman adds from admissions

Blackman's perspective applies the same logic to admissions reading: if a submission is not proportional to a student's standard of work, writing, or critical thinking, it stands out. The implication is that the method depends heavily on having a prior baseline. An instructor who knows a student well may spot the mismatch easily, while evaluators seeing an applicant for the first time may have less to compare against.

Practical Lessons for Instructors From Hanke's Approach

The defensive response described in the article focuses on making baseline comparisons easier and reducing opportunities for chatbot use.

  • Build a writing baseline early: Hanke's method depends on knowing a student's usual ability, so low-stakes, in-person writing samples early in a course give instructors a reference point for later work.
  • Change the assessment format: Teachers in the article have moved to AI-resistant or handwritten assignments to discourage chatbot-generated submissions on take-home work.