From Google to ChatGPT: How Student Cheating Patterns Are Evolving

Cheating in academics is nothing new. It likely dates back as long as schools have existed, but technology has dramatically changed how students might attempt shortcuts. The internet (think Google searches and copy-paste from websites) made it “easier than ever to find, share, and, now, produce answers rather than doing the work to solve problems independently” [1]. Today, generative AI chatbots like OpenAI’s ChatGPT (and rivals such as Anthropic’s Claude or Google’s Gemini) have become the new go-to cheating tools, capable of spitting out essays or solutions on demand. This marks a significant shift in cheating patterns, with implications for educators and students alike.
Note: Most of the studies quoted here have been conducted in the UK. This bias is owned by the simple fact that a lot of the serious research available about academic integrity in the AI-Era has been conducted there.

The Shift in Cheating Patterns: Plagiarism Down, AI Usage Up

In the pre-AI era, traditional plagiarism (copying existing text from online sources) was the number-one integrity concern. As recently as 2019, plagiarism accounted for nearly two-thirds of academic misconduct cases [2]. However, since the emergence of powerful AI text generators, the nature of cheating has changed. Universities are reporting surging AI-assisted cheating cases even as copy-paste plagiarism rates decline. The recent UK investigation found almost 7,000 proven instances of students misusing AI tools like ChatGPT in 2023-24 – over three times the rate of the previous year. In fact, one nationwide survey in the UK early 2025 showed that 92% of students use AI in some form, and 88% have used generative AI specifically for coursework, up from just 53% a year before [3]. Not all of that is outright “cheating,” of course, but tellingly 18% of students admitted directly including AI-generated text in their assignments – a clear academic integrity violation if done without permission.

Why are students gravitating to AI? Simply put, tools like ChatGPT can produce passable writing or solve problems in seconds, bypassing the need to do the work. “These tools enable students to generate assignments, essays, and problem solutions with minimal effort or understanding, thus circumventing traditional learning and assessment processes”, as one research study noted [4]. For a stressed or unprepared student, a chatbot can feel like an academic “fairy godmother for a last-minute essay deadline”, even as it becomes an educator’s nightmare [2]. And unlike a Google search that might lead a student to copy an existing source (which plagiarism detectors can catch), an AI model generates original text. This means conventional plagiarism checkers often fail to flag AI-produced work. It’s no surprise then that many students see AI as a tempting shortcut – and they’re even sharing tips online on how to avoid getting caught. Dozens of videos on TikTok for instance, now teach students how to use AI paraphrasing tools to “humanize” ChatGPT-written text and slip past AI detectors.

The Challenge of Academic Integrity in the AI Era

The rise of AI cheating has left educators and institutions scrambling to adapt. A major challenge is that AI-generated content is much harder to detect and prove compared to old-fashioned plagiarism. With copy-pasting, there’s usually a source that can be identified; by contrast, “in a situation where you suspect the use of AI, it is near impossible to prove” definitively [2]. AI detection software exists – for example, Turnitin’s anti-AI tool processed 130 million student papers and flagged 3.5 million as likely AI-written – but these tools are far from foolproof and and constantly produce false positives. Instructors have encountered cases of students being wrongly accused of using AI due to these imperfect detectors. The ‘cherry on top’ is, that AI detectors by design cannot deliver any actual proof [7] about their judgement. Understandably, faculty worry about unfairly penalizing honest students, so a mere suspicion of AI use can be tricky to act on. Dr. Peter Scarfe, who co-authored a study on AI and assessment, explains that AI-based cheating poses “a fundamentally different problem” than plagiarism and warns that the few students caught are likely just the tip of the iceberg. Indeed, his team was able to submit AI-written work into their university’s system without detection 94% of the time [5]. Students intent on cheating have noticed – hence the proliferation of those TikTok “how to evade AI detectors” guides. It’s an academic integrity arms race, and right now the playing field favors the savvy student using generative AI covertly.

Adapting to an AI-Enabled Academic World

How should educators respond to this new cheating landscape? There is growing consensus that purely punitive or traditional countermeasures won’t suffice. Banning AI tools outright or trying to “stuff the genie back in the bottle” is impractical (students will use them regardless, often undetected) [1], and it also ignores the potential benefits these tools can offer when used ethically. Instead, many experts suggest a dual approach: redesign assessments and embrace AI’s teachable uses. On one hand, teachers are rethinking assignments to make cheating more difficult – for example, by focusing on higher-order thinking, personal reflection, and unique analysis that AI can’t easily fake [4]. Routine fact-recall essays or generic prompts are being replaced with tasks that require critical reasoning, creativity, or personal context, which are much harder for ChatGPT to handle convincingly. Some instructors now require more in-class writing or oral presentations, so they can verify a student’s own understanding and voice [1]. Others incorporate personalized elements in assignments – e.g. asking students to tie concepts to their own experiences – making any AI-generated answers more obviously out of place [1]. In short, assessments are slowly shifting toward skills and outputs that can’t be easily outsourced to a bot.

On the other hand, completely shunning AI is seen as counterproductive, given that today’s students will likely use AI in their future workplaces. Many universities are therefore crafting clear policies on acceptable AI use rather than blanket bans. For instance, an instructor might allow using ChatGPT for brainstorming or editing help, but not for writing entire essays – and this is spelled out in advance. Educating students about when and how AI use crosses into cheating is key. “As educators, we need to clearly communicate to students what levels of AI use are permitted in our classrooms,” one teacher argues, since clarity can deter honest students from slipping into dishonest use [6]. By setting transparent guidelines and even teaching a bit about AI, teachers demystify the tools and emphasize that the goal is learning, not just getting answers.

Crucially, the motivation behind cheating also needs addressing. If students understand why an assignment matters and feel engaged in the learning process, they may be less tempted to outsource their work. “This all comes down to helping students understand why they are required to complete certain tasks” and designing assessments that actively involve them [2]. Professor Thomas Lancaster, an academic integrity researcher, suggests focusing on skills that AI cannot replace. For example, communication, interpersonal skills, and the ability to critically engage with new technology. Subsequently, students see genuine value in doing their own work. In other words, if assessments shift toward cultivating unique human skills and creativity, students will have fewer incentives to rely on AI shortcuts, and more incentives to use AI appropriately as a tool for learning.

Embracing the New Reality: Integrity and Innovation

Academic cheating has evolved from copying Wikipedia text to coaxing essays out of AI. This new reality is indeed challenging, but it’s also an opportunity to reform how we teach and assess. Educators and institutions are learning to balance integrity with innovation. They are updating honor codes and technologies and even collaborating with students to define ethical AI use. There is recognition that generative AI is here to stay in academia – much like calculators or the internet before it – and it can even enhance education if used the right way. As one policy guidance noted, “Generative AI has great potential to transform education… However, integrating AI into teaching, learning and assessment will require careful consideration and [schools] must harness the benefits and mitigate the risks” [2]. In practice, this means universities are investing in faculty training on AI, developing smarter assessment methods, and sharing best practices across departments. One of these assessment methods is introduced by Mentafy: Document the evolution of a text, rather than looking at the final product, which cannot be analyzed reliably enough anymore for academic integrity.

In conclusion, cheating patterns have undoubtedly changed – Google searches and copy-paste plagiarism are no longer the primary threat, superseded by AI-generated assignments. Yet the core mission remains the same: to uphold academic integrity and ensure students genuinely learn. By understanding the new cheating tools and thoughtfully adapting, educators can counter the misuse of AI while also guiding students towards honest, meaningful engagement with these powerful technologies. The challenge is significant, but a balance can be struck where AI shall become a partner in education rather than a menace, and where students choose learning over easy shortcuts. With vigilance, creativity, and open communication, the academic community can preserve trust and rigor in the age of AI.

Sources:

 [1]  Three Heads Blog (2024). AI and Academic Integrity: We Have Ideas.
 [2]  Goodier, M. (2025). Thousands of UK university students caught cheating using AI. The Guardian – Higher Education
 [3]  HEPI (2025). Student Generative AI Survey 2025 – Policy Note 61.
 [4]  Evangelista, E.D.L. (2025). Ensuring academic integrity in the age of ChatGPT. Contemporary Ed. Technology, 17(1).
 [5]  Scarfe, P. et al. (2024). A real-world test of artificial intelligence infiltration of a university examinations system: A “Turing Test” case study
 [6]  Dunlap, J. (2024). Cheating in the age of generative AI – High school survey (summary).
 [7]  Hiett R. (2025). AI Detectors: The Uses and the Risks in 2025