The way a story gets told shapes how people respond to it. And in early 2023, the dominant story about ChatGPT and higher education was a story about cheating and not about learning and opportunity. The focus was almost entirely on cheating. Sullivan, Kelly, and McLaughlan (2023) analyzed 100 news articles from Australia, New Zealand, the United States, and the United Kingdom to see exactly how that story was constructed, and their findings paint a picture that, three years later, looks like a textbook case of how panic becomes policy.
The articles were published between 2020 and February 2023, though the vast majority appeared after ChatGPT’s November 2022 release. This was the moment when generative AI went from a niche topic to a global conversation overnight, and the media was the primary channel through which the public, academics, and university administrators learned about it. What Sullivan et al. (2023) found in those articles tells us a lot about why so many institutions responded the way they did.
Cheating First, Learning Second
Of the 100 articles Sullivan et al. (2023) analyzed, 87 raised general concerns about cheating, and an equal number discussed strategies for discouraging students from using ChatGPT. The conversation was overwhelmingly tilted toward risk and prevention. Only 58 articles explored productive uses of ChatGPT in teaching, things like generating feedback, brainstorming, explaining complex concepts, debugging code, or creating sample assignments for classroom critique.
That gap between 87 and 58 matters. The media told the public that ChatGPT was primarily a threat to academic integrity, with learning potential as an afterthought. Sullivan et al. (2023) make the point clearly:
“Positioning the use of ChatGPT as a tool for cheating more often than a tool for learning can influence the perceptions that general readers have on the value of a university education, academic views on other institutional responses, and student thoughts on how such tools could be used in appropriate ways” (p. 35).
I covered Rudolph, Tan, and Tan’s (2023) much more provocative early take on ChatGPT, which leaned into the “bullshit spewer” framing. Sullivan et al. document the broader media ecosystem that produced and amplified that kind of response. The tone varied, with their sentiment analysis showing a roughly even split between positive (912 references) and negative (1,034 references) language. But the negative framing dominated the substantive conversations about policy and assessment.
The Policy Scramble
University responses, as Sullivan et al. (2023) tracked them through the news coverage, were fragmented and reactive. More institutions had banned ChatGPT (mentioned in 18 articles) than had explicitly allowed it (10 articles). But the most revealing number was 22: the count of articles describing universities as still reviewing or updating their policies. Indecision was the most common posture.
Where institutional policies didn’t exist, individual academics were creating their own rules course by course. Some moved back to pen-and-paper exams. Others proposed redesigning assessments to include oral presentations, podcasts, lab activities, vivas, and highly personalized prompts. There was no consensus about which disciplines or assignment types were most vulnerable to AI-generated work.
This tracks with what Moorhouse, Yeo, and Wan (2023) found when they analyzed actual assessment guidelines at top global universities in that same period. The policies were inconsistent, reactive, and often contradictory. Sullivan et al.’s media analysis adds another layer: the public conversation that was feeding those policy decisions was itself lopsided, framing the situation as a crisis of integrity with limited space for nuance.
The Equity Blind Spot
This is where Sullivan et al.’s (2023) findings get genuinely troubling. Only 10 of the 100 articles mentioned how ChatGPT could support students from disadvantaged backgrounds. Four noted its potential to reduce anxiety for students starting assignments. Three mentioned support for non-native English speakers. And just one article, a single article out of 100, referenced disability.
Sullivan et al. (2023) call this out directly: “ChatGPT also provides unique opportunities to enhance the academic success of students from different equity groups. This was not a common theme in the coding results and needed more discussion in the literature” (p. 36).
And the consequences are real: the institutions that were building their AI policies in early 2023 were doing so inside a media environment that barely mentioned equity. If the story you’re hearing is entirely about cheating prevention, your policy is going to be about cheating prevention. The students who stood to benefit most from AI tools, students with disabilities, students learning in a second language, students from under-resourced schools, were invisible in the conversation that shaped how universities responded.
Dawson, Bearman, Dollinger, and Boud (2024) later argued that the fixation on cheating had overtaken concerns about whether assessments were actually valid. Sullivan et al.’s data shows where that fixation came from: it was baked into the media framing from the very beginning.
Reading This in 2026
Sullivan et al.’s paper was published in March 2023, when ChatGPT was barely four months old. The media data they collected represents the very first public reaction to generative AI in education. And that reaction, as they document it, was overwhelmingly defensive: ban it, limit it, detect it, redesign assessments to avoid it.
Three years later, the story has shifted enormously. AI is embedded in mainstream tools. Most universities have moved past outright bans. The research base has grown from scattered early takes to a substantial body of empirical work. But the early media framing left marks. Institutions that built their first AI policies around fear and surveillance are the ones that had to rebuild from scratch when reality caught up. The cheating narrative didn’t just describe the moment. It shaped the decisions that followed, and some of those decisions cost universities years of productive experimentation.
The lesson for anyone following the AI-in-education conversation is simple: be careful whose story you’re listening to, and notice who isn’t in the room when the story gets told.
References
- Dawson, P., Bearman, M., Dollinger, M., & Boud, D. (2024). Validity matters more than cheating. Assessment & Evaluation in Higher Education, 49(7), 1005–1016. https://doi.org/10.1080/02602938.2024.2386662
- Rudolph, J., Tan, S., & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of traditional assessments in higher education? Journal of Applied Learning & Teaching, 6(1), 342–354. https://doi.org/10.37074/jalt.2023.6.1.9
- Sullivan, M., Kelly, A., & McLaughlan, P. (2023). ChatGPT in higher education: Considerations for academic integrity and student learning. Journal of Applied Learning and Teaching, 6(1), 31–40.