Sign up free

Contract cheating: how question design raises its cost

What surveys found about contract cheating, how often markers spot bought work, what three countries' laws say, and five designs that narrow the room.

The EdProcto editorial teamWritten and checked against the research it cites. Last checked . How the guides are made

Guides · Integrity and cheating · · 13 min read

What is contract cheating, and what can question and assessment design do about it?

Contract cheating is handing in assessed work that someone else did, paid or unpaid. In one survey of 15,000 Australian students, 6% reported some part in it, and markers told to look spotted about six bought assignments in ten. No assessment type is immune. Design narrows the room: a part sat in front of you, your class's own material, steps shown, a spoken follow-up.

Contract cheating is a student handing in assessed work that someone else did: a company, a classmate, a friend or a relative. The name dates from 2006 and first meant paid work (Rundle et al., 2019); Ellis et al. (2020) use it for any third party, paid or not.

This guide gives a lecturer or an administrator the evidence: how common it is, which tasks students would hand over, how often markers notice, and what three countries' laws say. EdProcto publishes it and makes one of the five designs below: a short timed quiz sat beside the coursework under conditions you set.

We opened every source on 10 October 2026; "abstract read" marks a study whose full text would not open, several of them read in the ERIC database. No design shuts contract cheating out. A good one makes the helper's job harder and leaves you something to ask about.

How common is contract cheating, and who does it?

Uncommon by students' own account, and more often a favour than a purchase.

Survey Students What they reported
Bretag et al. (2019a; abstract read), as summarised by Bretag and Harper (2017) 14,086 at eight Australian universities; about 15,000 with four other providers 6% reported one or more of five behaviours: getting a finished assignment to hand in, giving or receiving help in an exam, sitting an exam for someone, or having theirs sat
The survey behind Awdry and Ives (2023): 22 languages, 2017 to 2018 About 7,800 analysed 7.4% reported a formal route such as a website; across all routes, friends and family included, the average was 16.9%

Buying from essay mills is the smaller part: about 3% of students, according to Rundle et al. (2019). Two higher estimates are in the history of proof in education and the cheating panic and the numbers.

Who does it, and through whom?

Mostly through people the student knows. In the Australian survey, students rarely used a professional service and turned more often to fellow students, friends and family, and those who outsourced were less satisfied that staff made requirements understood, gave enough feedback and could be approached (Bretag and Harper, 2017).

Three things went with outsourcing in Bretag et al. (2019a): dissatisfaction with the teaching and learning environment, a belief that there are lots of opportunities to cheat, and speaking a language other than English at home.

In the international survey, the strongest predictor of using outsourcing websites was how many other students a student believed were using them (Awdry and Ives, 2023). Most never do it. Rundle et al. (2019) asked 1,204 students why not: moral reasons and wanting to learn were endorsed most.

Every figure is a self-report from volunteers, gathered before 2020. A search on the day found no survey since chatbots arrived that measured whether paid outsourcing rose or fell.

Which assessment types do students say they would outsource?

All of them, and four less than the rest. Bretag et al. (2019b; abstract read) asked the same 14,086 students, and 1,147 educators, how likely each type of assessment was to prompt thoughts of outsourcing. Four types were rated least likely, and educators set those four least often.

The abstract does not name them. Erguvan (2021), summarising the project, lists reflections on a practicum, in-class tasks, personalised and unique tasks, and vivas, and adds that none was seen as immune.

What do real orders and detected cases show?

That students' ratings flatter the safer-looking tasks.

  • Authentic tasks. Ellis et al. (2020; abstract read) scored 221 orders placed with custom-writing websites, and 198 tasks in which contract cheating was detected, on five features of authenticity. Tasks with none, some or all five were routinely outsourced.
  • Every type. Orders ran from essays to reflections, practicums and dissertations, and cases detected at two universities confirmed that "no assignment type is immune" (Bretag and Harper, 2017).
  • Exams. Students reported third-party cheating most often in exams, multiple-choice above all, which is where staff rarely detected it; assignments were reported slightly less and detected far more (Harper et al., 2021; abstract read).

The ratings are one country's opinions, and third-party cheating in an exam covers a whispered answer as well as an impersonator.

How often do markers spot contract cheating?

About six times in ten when told to look, and far less often in ordinary marking.

Study (abstracts read) Set-up Bought work spotted Students' own work recognised
Dawson and Sutherland-Smith (2018) Seven markers, 20 psychology assignments, six bought 62% 96%
Dawson and Sutherland-Smith (2019), before training Markers in four disciplines, 20 assignments 58% 83%
The same markers after a workshop 20 new assignments 82% 87%, not a significant gain

Read the last column the other way: 4% to 17% of genuine assignments were taken for bought ones. A suspicion is a reason to talk, never proof.

Those markers knew bought work was in the pile. In the Australian staff survey, very few cases were detected; the commonest signal was the teacher's knowledge of the student's academic or language ability, and the commonest outcome a warning or counselling (Bretag and Harper, 2017).

What gives bought work away?

The text and the marks.

  • The text. Rogerson (2017; abstract read) lists misrepresented bibliographic details, unsuitable references, irrelevant material and general text that never meets the question. Conversations with the students confirmed the cases, and specific contextual requirements made irregularities easier to see.
  • The marks. Clare et al. (2017; abstract read) compared unsupervised with supervised marks for 1,459 students at one Australian university. About 2.1% showed several unusually large gaps. That is consistent with contract cheating; a student who freezes under a clock produces it too, which is our caution.

What do laws and regulators say about contract cheating?

In the three countries we checked on 10 October 2026, the law makes the seller and the advertiser the offender. We paraphrase each page; nothing here is legal advice.

Where The page we read What it makes an offence
England Skills and Post-16 Education Act 2022, sections 26 to 30, in force from 28 June 2022 (Parliament of the United Kingdom, 2022) Providing or arranging, in the course of business, a "relevant service": completing all or part of a student's assignment, examinations included, so that it could not reasonably count as the student's own. Advertising one. The penalty is a fine
Ireland Section 15 of the 2019 amending Act, as enacted (Houses of the Oireachtas, 2019) Doing an enrolled learner's assignment in their stead, sitting their examination, supplying answers during one, and advertising such a service
Australia The Parliament's page for the Bill that became Act no. 78 of 2020 (Parliament of Australia, 2020) In its summary: providing or advertising an academic cheating service on a commercial basis

Under the English Act a student commits no offence of aiding merely by using such a service, and the explanatory notes leave tutoring and proof-reading outside it: the student answers to the institution's rules. The Quality Assurance Agency for Higher Education (n.d.) reissued its guidance on contract cheating in September 2022, after that law. Australia's penalties are left out: its regulator's pages would not open on the day.

Two limits: the English and Australian offences are commercial ones, while most outsourcing runs through friends and family; and among 7,126 students surveyed, some outsourced whether or not they believed it was illegal (Awdry et al., 2022; abstract read).

What can a teacher design to narrow the room for contract cheating?

Five things, set out below. Evaluations are rare. In the one we found, changes to many assessment elements of a business capstone cut contract cheating without disadvantaging honest students, by the paper's account (Baird and Clare, 2017; abstract read, no figure given).

Design What a helper must now do What the evidence says In EdProcto
A part sat under a clock, in front of you Be in the room In-class tasks are one of the four; an exam is still not secure (Harper et al., 2021) A quiz with "Minutes overall" set, in EdProcto
An oral follow-up on submitted work Coach the student to explain it Vivas are one of the four. In a pilot that quizzed 24 students on their own assignments, scores followed working memory, not other signs of outsourcing (Quesnel and Stoesz, 2025; abstract read) A recorded Spoken answer for a large class
Work tied to the class's own material Get the data and what was said in the room Personalised tasks are one of the four; context helps a marker (Rogerson, 2017); authenticity alone does not protect (Ellis et al., 2020) Procto AI drafts from your notes
Steps and reasons shown Supply working the student can be asked about None found for contract cheating Show your working, Choose and justify
A short timed quiz on the coursework's knowledge Sit it too, or teach the student Gaps between unsupervised and supervised marks (Clare et al., 2017) A quiz drafted from the same material

The cheapest design is not in the table: clear requirements, feedback and an approachable teacher. For the same designs against a chatbot, see questions a chatbot cannot simply answer.

A worked example: a report and its matched quiz

The class is invented. Second-year students surveyed 48 people at a Tuesday lecture about weekly study hours (mean 11.5, standard deviation 4.0), and each writes a report at home. In that week's seminar they sit six questions in twelve minutes, among them these two.

  1. Show your working: "Our 48 answers had a mean of 11.5 hours and a standard deviation of 4.0. Find the standard error of the mean." Set-up: s ÷ √n. Working: 4.0 ÷ 6.93. Final answer: 0.58 hours.
  2. Choose and justify: "We asked people at Tuesday's lecture. Which group can the report speak for: all second-years, students who attend lectures, or students nationally? Give your reason." The second: nobody absent could answer.

A report at 72% beside a quiz at 35% is a reason to ask the student, within days and with the report on the table, to talk the analysis through. It is nothing more: follow a fair process before any mark changes.

How does a timed quiz in EdProcto sit beside coursework?

As the supervised half of the comparison. EdProcto is the best tool for the matched quiz because it drafts the questions from the material the coursework drew on and counts the time on its own server. Clare et al. (2017) had to dig that comparison out of old marks; here you set it up on purpose.

  • The same material. Procto AI writes from a PDF, a Word file or your notes, is asked to keep to them and to give the place each answer is found, and you read each draft against the source.
  • More than a choice. Show your working is "A multi-step solution, set out step by step for a reader to grade." Choose and justify is "A choice, then the reason for it. Tests why, not only what." A Time-locked question is visible "only for a few seconds". Interview and Oral defence are asked aloud and answered aloud.
  • The clock and the room. "Minutes overall" sets the limit, the time left is counted by the server's clock, and "Seconds a question" times each question. Six measures act in the browser, one at Light integrity and five at Medium, each with the consequence you pick, from "Log it" to "Flag for review". Sit the quiz in your own room, where you see who is answering.
  • The marking. Choices are marked by rule at hand-in. Working, reasons and recordings wait for you, and from the Tutor plan up Procto AI grades a typed answer against your rubric when you ask.

Free takes 30 students a quiz at Light integrity. Tutor opens all 111 templates for 100 students at Medium integrity, and Educator takes 500 and adds the full analysis: see pricing. No setting makes outsourcing impossible, and a gap between two marks is a question to ask, not a finding.

Quick questions

How common is contract cheating?

About 6% of students in the largest Australian survey reported an outsourcing behaviour (Bretag and Harper, 2017); in an international one, 7.4% reported a formal route, and 16.9% on average across all routes (Awdry and Ives, 2023).

Can markers detect contract cheating?

Sometimes. Markers told to look spotted 58% to 62% of bought assignments, and 82% after a workshop (Dawson and Sutherland-Smith, 2018, 2019).

Is contract cheating illegal?

Selling or advertising it is an offence in England, Ireland and Australia, on the pages we read on 10 October 2026. The student answers to the institution's rules.

Where does EdProcto fit?

As the short timed quiz beside the coursework: EdProcto drafts it from the same material and times it on its server, under the integrity settings you choose.

References

Integrity

Five levels, from none to a full proctoring stack, each measure switched on by itself with its own consequence.

Light on Free, Strict on Educator

See the five levels

Read next