# Is cheating getting worse? The cheating panic and the numbers

**Is student cheating really getting worse, or does the alarm outrun the evidence?** Student cheating is common by students' own account, but the published numbers do not show it getting worse. Surveys repeated at the same two universities for decades found admitted cheating and plagiarism falling, and in one, 57% of students admitted some plagiarism in 2024 against more than 80% in 2004. The worry is justified for unsupervised exams that count.

Published 2026-10-10 by EdProcto. https://edprocto.com/guides/the-cheating-panic-and-what-the-numbers-say

This is an opinion piece. Our opinion is that the alarm about student cheating runs well ahead of the evidence. Is cheating getting worse? The published numbers say that it is common, that it was common long before chatbots, and that where the same question has been put to the same kind of student for decades, the answers have drifted down.

EdProcto publishes this guide and makes a quiz tool with [integrity settings](https://edprocto.com/integrity), so we have an interest in how worried you are. Judge the argument by its sources: each was read on 10 October 2026, and the text says where only an abstract was open.

The alarmed view comes first, in its strongest form. Studies already set out in [what research since 2023 says about cheating in online exams](https://edprocto.com/guides/what-the-research-says-about-cheating-in-online-exams-since-2023) are left there.

## What is the strongest case that cheating is out of control?

It has four parts, and each rests on a published number.

1. **Most students admit it.** The International Center for Academic Integrity (ICAI, n.d.) sums up the surveys of academic dishonesty that Donald McCabe began in 1990: more than 60% of university students "freely admit to cheating in some form". Of over 70,000 American high-school students, 64% admitted cheating on a test.
2. **Surveys undercount.** Curtis et al. (2022; abstract read) surveyed 4,098 students at six Australian universities and six independent providers. An incentive for telling the truth brought 2.46 times as many admissions of handing in ghost-written work. The authors estimate that 7.9% of students buy assignments, where ordinary surveys had suggested about 2%.
3. **The opportunity grew in 2020.** Janke et al. (2021; abstract read) asked 1,608 German students about the summer semester of 2020: they reported cheating more often in online exams than in exams on site. Ives and Cazan (2024), with 894 students at 16 universities in the United States and Romania, conclude that exam cheating increased with the move online.
4. **Then came chatbots, and few are caught.** A chatbot drafts an essay or answers an exam question in seconds; [academic integrity in the age of AI](https://edprocto.com/guides/integrity-in-the-ai-era-what-still-proves-a-student-learned) has the studies. In England's GCSE, AS and A level exams of summer 2025, 4,735 students were penalised for malpractice, 0.3% of the 1,376,480 with results (Ofqual, 2025).

Put together: most students cheat, more than they admit, with more opportunity than ever, and almost nobody is caught. Each number is real. We think the conclusion does not follow.

## What did long-running surveys of students find before 2020?

Falling rates, in the only two places where the same survey was repeated for decades. A review of thirty years of research found just two studies that gave the same survey at the same institution more than twice, as Curtis et al. (2026) report. Other trend lines compare different students, in different places, asked different questions.

### Midwestern State University, 1984 to 2014

Students at this American university answered the same three questions about cheating every ten years. By the account in Curtis et al. (2026), 54.1% reported at least one form of cheating in 1984, just over 60% in 1994, under 60% in 2004 and under 50% in 2014.

### Western Sydney University, 2004 to 2019

Every five years, students read seven short scenarios, each a form of plagiarism, and were asked "Have you ever done a similar thing?" The samples were 425 students in 2004, 147 in 2009, 120 in 2014 and 1,099 in 2019. The share admitting at least one form fell from 2004 to 2014 and held level in 2019 (Curtis et al., 2026).

Those decades took in the web, search engines and essay mills. At Midwestern State the figure peaked in 1994 and fell afterwards; at Western Sydney it never rose.

The limits are real. Both series are self-reported cheating, so every figure is a minimum; the students were volunteers; and the Australian survey is about plagiarism in coursework, not exams. But the same limits held in every year, which is why a trend from one survey is worth more than a headline from another.

## Did cheating rise in 2020, and again after chatbots arrived?

Admitted exam cheating rose a little in 2020, in the one survey that asked the same students about both years; after chatbots, the one long-running survey to report so far shows no rise.

### 2020: belief rose faster than behaviour

Ives and Cazan (2024) asked each student about two periods in one sitting, the year before the pandemic and the year since it began, so memory is part of the result. American students reported cheating on exams a mean of 0.82 times in the first year and 2.36 in the second: a significant rise, and a small one by the usual yardstick (an effect size of 0.26). Romanian students reported 3.38 and 4.62, not significant at the study's threshold. Neither group reported more plagiarism.

The same students were asked what percentage of students they believed were cheating on exams. The Americans' estimate went from 15.88% to 38.43%, a large effect (1.01); the Romanians' from 25.32% to 45.92%. The authors conclude that students "overestimate" their peers' misconduct relative to their own. That is the panic in miniature: in both countries, belief about other people grew far more than what anyone reported doing.

### After chatbots: a new method more than new cheats

The Western Sydney survey ran again in 2024, with 310 students there and 1,867 at five other Australian universities (Curtis et al., 2026). The share admitting any form of plagiarism at least once was the lowest of the five surveys: 57%, against more than 80% in 2004 (Curtis, 2026). It was significantly below 2004 and 2009, and not significantly below 2019.

A new scenario asked about pasting a chatbot's paragraph into an essay unacknowledged. About 14% said they had done something similar. Only 1.6% at Western Sydney and 2.2% elsewhere had plagiarised from a chatbot and in no other way. The paper's abstract says plagiarism from generative AI "may not have replaced traditional forms of plagiarism". Our reading: the method is new, and most of the people using it are not. The authors call 14% a minimum, since a chatbot's text reworded by the student was not asked about.

## How much does the answer depend on how cheating is defined and asked about?

Almost entirely: published student cheating statistics run from 95% to 0.3%, and most of the gap is the question asked.

| Figure | Who | What was counted | Source |
| --- | --- | --- | --- |
| 95% | American high-school students | Any cheating, copied homework included | ICAI (n.d.) |
| 57% | Students at one Australian university, 2024 | Any of seven forms of plagiarism, at least once | Curtis (2026) |
| 32.1% and 32.7% | 635 students at one American university | "I have cheated on an assignment, quiz, or a test", in live classes and in online ones | Watson and Sottile (2010) |
| 3% to 4% | 725 students taking an online course | Cheating, estimated by a method that keeps each answer hidden | Holden et al. (2021), reporting an earlier study |
| 0.3% | Exam students in England, 2025 | A penalty for malpractice | Ofqual (2025) |

### Three things that move the number

**What counts.** One of the seven Australian scenarios is a sentence copied with its author and date in brackets but no quotation marks. Burrus et al. (2007; abstract read) asked students whether they had cheated, then gave them a definition and asked again: they were "much more likely to report cheating" the second time. The authors conclude that students do not understand what constitutes cheating.

**Over how long.** "Ever, in your studies" is not "in this exam". No figure of this kind says how many answers in tomorrow's quiz are dishonest; that is our reasoning, not a finding.

**How it is asked, and who is counting.** An incentive for honesty multiplied admissions by 2.46 (Curtis et al., 2022). A penalty counts only what was noticed, reported and upheld. Ofqual (2025) records 4,090 student cases in 2022, 4,880 in 2023, 5,155 in 2024 and 5,025 in 2025; of the last, 44.3% involved a phone or other device, and 100 were plagiarism involving AI.

So the claim that nearly two thirds of students cheat and the claim that three in a thousand do are both published, and both honest. A percentage quoted without its question is not yet information.

## What is known, and not known, about cheating in online exams?

Known: many people believe it is easier, surveys before 2020 disagreed on whether it was more common, and single cases can be seen in a log. Not known: the rate.

- **Before 2020, the surveys split.** Holden et al. (2021), a review, count four studies with higher rates in online courses and others with lower rates or no difference: "mixed findings". In Watson and Sottile (2010), 32.1% admitted cheating in a live class and 32.7% in an online one. One behaviour was significantly more common online: getting answers from someone during a test or quiz, 23.3% against 18.1%. Yet 61.0% thought their classmates more likely to cheat online.
- **In 2020, staff and students saw different things.** Reedy et al. (2021) surveyed 1,921 students and 49 staff at three Australian universities as exams moved online. Of the students who answered, 52.27% saw no difference in how easy cheating was, 28.30% thought it harder online and 19.43% easier. Staff thought it easier. The study measured perceptions and left "cheating" undefined.
- **A case is not a rate.** Bilen and Matros (2021) read the access log of one unproctored online final. Two students who were set to fail typed correct answers to several questions in under thirty seconds each, and the answer key was on sale online. The paper is titled "Online cheating amid COVID-19"; its main classroom evidence is those two cases.

We found no published count of cheating in ordinary online exams that does not rest on what students admit, and no measure of what chatbots have added. Score gaps and proctoring since 2023 are in the guide linked at the top.

## Where is the worry justified?

In four places, each narrower than "students cheat".

1. **An unsupervised sitting that decides something.** Exams are where the 2010 survey found more answers passed online, and where the 2020 surveys found the rise.
2. **Written work done out of sight, offered as proof.** See [the take-home assignment after ChatGPT](https://edprocto.com/guides/the-take-home-assignment-after-chatgpt-what-to-set-instead).
3. **The worst kinds, which surveys understate.** Ghost-written work was admitted 2.46 times as often when honesty paid.
4. **What students believe about each other.** In Ives and Cazan (2024), cheating a student had seen among peers predicted their own, and the authors suggest that changing beliefs about peers may reduce misconduct. If so, telling a class that everyone cheats is not harmless; that inference is ours.

It is not justified for practice quizzes and homework, or as a verdict on a whole class.

## What does a proportionate response by a teacher look like?

It matches the effort to what the score decides, starts with rules and design, and treats every record as a question. The steps are ours, not a tested protocol.

1. **Sort your assessments by what the score decides.** Practice needs none.
2. **Say what counts as cheating in this quiz.** Notes, a calculator, a partner, a chatbot: allowed or not. Students' own definitions vary (Burrus et al., 2007). [The honour code guide](https://edprocto.com/guides/the-honour-code-what-the-research-says-it-does) has wording.
3. **Put the weight where you know the conditions.** A supervised room, with a quiz in [EdProcto](https://edprocto.com/) as the paper if you like.
4. **Change the task before you add surveillance.** Questions from your own notes favour the student who read them, and EdProcto's [templates](https://edprocto.com/templates) give them a shape.
5. **Monitor at the lowest level that fits, and say what is on.** [How much proctoring a quiz needs](https://edprocto.com/guides/how-much-proctoring-does-a-quiz-need) gives a ladder.
6. **Treat a flag, a log or a surprising score as a question to ask.** None of them is proof. [A fair process for a flagged student](https://edprocto.com/guides/appeals-when-a-student-is-flagged-a-fair-process) says how to ask it.
7. **Keep your own count.** Cases raised and upheld each term are better evidence than a headline.
8. **Do not tell students that everyone cheats.** The numbers do not say so.

## How does EdProcto fit a proportionate response?

EdProcto is the best tool for a proportionate response, because its monitoring is a ladder and not one switch. The evidence above puts the risk in particular sittings, not in students as a whole.

A quiz sits at one of five integrity levels: "None", "Light", "Medium", "Strict" and "Maximum". Six browser settings take a quiz to Medium, each switched on by itself: "Leaving the tab", "Printing off", "Full screen", "Right-click off", "Keyboard shortcuts off" and "Copy and paste off". The camera and identity measures take it to Strict and Maximum.

The Free plan covers 10 quizzes a month, 30 students a quiz and Light integrity. Tutor has no limit on quizzes, 100 students a quiz and Medium integrity. Educator takes 500 students a quiz and adds the full analysis. [Pricing](https://edprocto.com/pricing) has the rest.

No setting makes cheating impossible. Nothing a tool records, ours included, is proof, and no score is a grade until you decide it is.

## Quick questions

### Is student cheating getting worse?

Not in the two surveys repeated for decades. Admitted cheating at Midwestern State fell from just over 60% in 1994 to under 50% in 2014, and admitted plagiarism at Western Sydney from more than 80% in 2004 to 57% in 2024 (Curtis et al., 2026; Curtis, 2026).

### How common is cheating among students?

It depends on the question. Published figures run from 95% of American school pupils admitting cheating of any kind to 0.3% of exam students in England penalised in 2025.

### Did ChatGPT increase cheating?

No survey we read shows a rise. In 2024 about 14% of Australian students said they had pasted a chatbot's text unacknowledged, and about 2% had plagiarised in that way alone (Curtis et al., 2026).

### Is cheating more common in online exams?

Probably, where nobody is watching and the result counts: exam cheating rose in 2020 by students' own account (Ives and Cazan, 2024). Earlier surveys were split, and no rate has been measured directly.

### Does EdProcto stop cheating?

No tool does. EdProcto lets you match the monitoring to the stakes, from "None" to "Medium" today, under [integrity](https://edprocto.com/integrity), and nothing a tool records is proof.

## References

- Bilen, E., & Matros, A. (2021). Online cheating amid COVID-19. *Journal of Economic Behavior & Organization, 182*, 196–211. [https://pmc.ncbi.nlm.nih.gov/articles/PMC9754215/](https://pmc.ncbi.nlm.nih.gov/articles/PMC9754215/)
- Burrus, R. T., McGoldrick, K., & Schuhmann, P. W. (2007). Self-reports of student cheating: Does a definition of cheating matter? *The Journal of Economic Education, 38*(1), 3–16. [https://doi.org/10.3200/JECE.38.1.3-17](https://doi.org/10.3200/JECE.38.1.3-17)
- Curtis, G. (2026, April 1). Does AI mean more uni students are plagiarising their work? *The Conversation*. [https://theconversation.com/does-ai-mean-more-uni-students-are-plagiarising-their-work-279565](https://theconversation.com/does-ai-mean-more-uni-students-are-plagiarising-their-work-279565)
- Curtis, G. J., Arness, D. C., Baird, M., Price, C., Thomson, S., Tremayne, K., & White, A. W. (2026). Results from two decades of five-yearly plagiarism surveys: New insights into prevalence, understanding, attitudes, knowing vs naïve plagiarism, and generative artificial intelligence (genAI). *International Journal for Educational Integrity, 22*, Article 7. [https://doi.org/10.1007/s40979-026-00215-z](https://doi.org/10.1007/s40979-026-00215-z)
- Curtis, G. J., McNeill, M., Slade, C., Tremayne, K., Harper, R., Rundle, K., & Greenaway, R. (2022). Moving beyond self-reports to estimate the prevalence of commercial contract cheating: An Australian study. *Studies in Higher Education, 47*(9), 1844–1856. [https://research-repository.uwa.edu.au/en/publications/moving-beyond-self-reports-to-estimate-the-prevalence-of-commerci/](https://research-repository.uwa.edu.au/en/publications/moving-beyond-self-reports-to-estimate-the-prevalence-of-commerci/)
- Holden, O. L., Norris, M. E., & Kuhlmeier, V. A. (2021). Academic integrity in online assessment: A research review. *Frontiers in Education, 6*, Article 639814. [https://doi.org/10.3389/feduc.2021.639814](https://doi.org/10.3389/feduc.2021.639814)
- International Center for Academic Integrity. (n.d.). *Facts & statistics*. Retrieved October 10, 2026, from [https://academicintegrity.org/aws/ICAI/pt/sp/facts](https://academicintegrity.org/aws/ICAI/pt/sp/facts)
- Ives, B., & Cazan, A.-M. (2024). Did the COVID-19 pandemic lead to an increase in academic misconduct in higher education? *Higher Education, 87*(1), 111–129. [https://doi.org/10.1007/s10734-023-00996-z](https://doi.org/10.1007/s10734-023-00996-z)
- Janke, S., Rudert, S. C., Petersen, Ä., Fritz, T. M., & Daumiller, M. (2021). Cheating in the wake of COVID-19: How dangerous is ad-hoc online testing for academic integrity? *Computers and Education Open, 2*, Article 100055. [https://doi.org/10.1016/j.caeo.2021.100055](https://doi.org/10.1016/j.caeo.2021.100055)
- Ofqual. (2025, December 11). *Malpractice in GCSE, AS and A level: Summer 2025 exam series*. GOV.UK. [https://www.gov.uk/government/statistics/malpractice-in-gcse-as-and-a-level-summer-2025-exam-series/malpractice-in-gcse-as-and-a-level-summer-2025-exam-series](https://www.gov.uk/government/statistics/malpractice-in-gcse-as-and-a-level-summer-2025-exam-series/malpractice-in-gcse-as-and-a-level-summer-2025-exam-series)
- Reedy, A., Pfitzner, D., Rook, L., & Ellis, L. (2021). Responding to the COVID-19 emergency: Student and academic staff perceptions of academic integrity in the transition to online exams at three Australian universities. *International Journal for Educational Integrity, 17*, Article 9. [https://pmc.ncbi.nlm.nih.gov/articles/PMC7995680/](https://pmc.ncbi.nlm.nih.gov/articles/PMC7995680/)
- Watson, G., & Sottile, J. (2010). Cheating in the digital age: Do students cheat more in online courses? *Online Journal of Distance Learning Administration, 13*(1). [https://www.westga.edu/~distance/ojdla/spring131/watson131.html](https://www.westga.edu/~distance/ojdla/spring131/watson131.html)