When AI helps with coursework
How exam boards set rules for declaring AI use and for checking that coursework is the student's own
IAPQB Report No. 13 · Rule explainer · Published 17 November 2026
International AI Practice Qualification Board (IAPQB)
In one minute
- In the UK, guidance from the Joint Council for Qualifications (JCQ) allows AI use in assessed work if it is properly acknowledged, and gives no credit for work that is not the student's own. A new version took effect in September 2026 [1].
- The International Baccalaureate (IB) will not ban AI software, and requires it to be credited in the text and the bibliography [2].
- In England, penalised cases of plagiarism involving AI at GCSE, AS and A level rose from 85 in 2024 to 100 in 2025. There were 5,025 student malpractice cases in 2025 [3].
- The boards rely on records and on teachers who see work as it develops [1, 4]. In two of the US College Board's Advanced Placement (AP) courses, a missed checkpoint means a score of 0 on the associated task [4].
1. A familiar situation
A teacher has coursework to sign off. One essay is far more polished than the student's earlier work, and a detection tool says AI probably wrote it. The student says a tool only helped with wording and a translation. Nobody kept a record. The teacher's signature will tell the exam board that the work is the student's own.
2. Why it happens
Coursework is often written outside the classroom, where no one sees every step. Ofqual, the regulator in England, says supervised exams remain largely protected, while non-exam assessment, mainly coursework done outside controlled conditions, faces greater pressure [5]. In College Board surveys, the share of high school students who report using generative AI for schoolwork rose from 79% to 84% between January and May 2025 [6]. JCQ says research indicates that "AI use cannot be reliably identified from the text alone". It also warns that detection tools can produce false positives and false negatives [1].
The IB calls a ban "an ineffective way to deal with innovation", and requires AI software to be "credited in the body of the text and appropriately referenced in the bibliography" [2]. JCQ says: "When properly acknowledged and referenced, AI use is permitted", although students cannot be credited for work that is not their own [1].
IAPQB view. Each board here asks for evidence it can check beside the finished work. JCQ wants a dated acknowledgement with saved prompts, the IB wants credit in the text and bibliography, and the College Board wants checkpoints seen by a teacher [1, 2, 4]. JCQ presents all of its anonymised penalty examples as misuse, which it defines as AI use that was not appropriately acknowledged in work that is not the student's own [1].
3. Three cases
3.1 United Kingdom, 2026: the shared rulebook is rewritten
JCQ's guidance names contacts at seven awarding bodies, including AQA, Cambridge OCR, Pearson and WJEC [1]. After Revision 2 was published on 30 April 2025 [7], a newer version took effect in September 2026 [8, 1].
The 2026 version defines AI misuse as using AI tools without appropriate acknowledgement and submitting work that is not the student's own. This is malpractice, with sanctions from a warning to disqualification [1]. Among the changes in this version, students must receive a JCQ document, Information for Candidates: AI, before non-exam assessment starts. Staff must prohibit concealed or wearable technology such as smart glasses, and an AI tool cannot be the sole marker of student work [1].
In one anonymised example, research translated with the AI tool DeepL appeared unacknowledged in an A level Art and Design sketchbook, and the component received zero marks. Penalties in the examples ranged from loss of marks for one section to disqualification [1].
3.2 England, 2024 to 2026: AI cases are counted, and the regulator asks for more
Ofqual's yearly malpractice figures for GCSE, AS and A level in England have shown plagiarism involving AI separately since 2024 [9]. In summer 2025, plagiarism involving the misuse of AI accounted for 100 cases: 75.0% of student plagiarism cases and 2.0% of all student malpractice cases. In summer 2024 the figures were 85 cases, 55.4% and 1.7% [3].
On 16 July 2026 Ofqual updated its approach to AI. It says "the number of detected cases remains relatively low", while teachers and leaders are significantly concerned about how widespread misuse could be. The steps it says it has already taken include "writing to exam boards to require stronger deterrence, detection and authentication measures" [5].
3.3 United States, 2026-27: the College Board ties AI use to checkpoints
The College Board's AI guidance for Advanced Placement (AP) courses in 2026-27 sets rules course by course [4]. In AP Seminar and AP Research, generative AI is an optional aid, for example to explore topics or check grammar and tone. Students must still read sources directly and do their own analysis. Teachers must affirm, to the best of their knowledge, that students completed the checkpoints authentically. Failure to complete them results in "a score of 0 on the associated task" [4]. In AP Art and Design, generative AI may not be used to create final submitted artwork, and any AI use must be disclosed in the student's portfolio [4].
4. How practitioners handle it now
The rules are set out before the work starts. JCQ says a school's malpractice policy must include AI, and that teachers should be clear about when and if students may use AI tools [10]. Students must receive Information for Candidates: AI before non-exam assessment begins [1].
Students name the tool and keep the evidence. JCQ requires the name of the AI source and the date the content was generated. The student keeps the prompts and output in a non-editable form, such as a screenshot, adds a short explanation, and submits them with the work. A reference to "AI" or "ChatGPT" alone is not acceptable [1]. The IB requires credit in the text and in the bibliography [2].
Teachers and students sign declarations. Under JCQ's instructions, teachers sign a declaration of authentication after the work is completed. Centres record a mark of zero if a candidate cannot confirm that the work is authentic [1]. A concern found before the student signs follows the school's malpractice policy. After signature, it must be reported to the awarding body [1, 10].
Teachers see the work while it is being done. JCQ advises talking to students about their work on an ongoing basis, setting some work in class where possible, and comparing new work with earlier work [10]. The College Board's checkpoints make this a scoring rule [4].
Detection software is one input among several. JCQ says detection tools must be one part of a wider judgement about authenticity, and warns that they can produce false positives and false negatives [1]. In four of its five anonymised examples, a teacher, examiner or assessor raised the concern first, and software or other checks followed [1].
5. What to take from this
For students and others who submit assessed work. A JCQ acknowledgement needs the tool's name, the date, the saved prompts and output, and a short explanation [1]. JCQ and the IB both treat AI-made content as work that is not the student's own, so it earns little or no credit even when acknowledged [1, 2, 10]. The DeepL example shows that AI translation tools are covered [1].
For teachers and assessors. In the JCQ process, the student's signed declaration separates a school's own handling of a concern from a formal report to the awarding body [1, 10]. Because JCQ warns of false results, drafts and conversations seen during the work give firmer grounds for a teacher's signature than a detection score alone [1, 10].
For schools, training providers and employers. Together, these rulebooks use four parts: a written rule on allowed AI use, a record of that use, a signed declaration, and contact with the assessor during the work [1, 2, 4]. In England the rules have not ended the regulator's concern: by July 2026 Ofqual had written to exam boards to require stronger deterrence, detection and authentication measures [5].
6. Limits of this report
Other exam boards and countries, not covered here, have their own rules. The IB statement shows no date, and the IB's academic integrity policy could not be opened. JCQ's AI web page still describes the April 2025 revision; this report uses the September 2026 version [7, 8, 1]. Ofqual's figures cover only GCSE, AS and A level in England, and only penalised cases [3, 9]. Figures for summer 2026 are due on 10 December 2026, a date Ofqual has confirmed [11]. JCQ's five examples cannot show how often each pattern occurs. No detection tool was tested.
References
[1] Joint Council for Qualifications (2026). AI Use in Assessments: Your role in protecting the integrity of qualifications. Effective from September 2026. JCQCIC. https://www.jcq.org.uk/wp-content/uploads/sites/2/2026/10/AI-Use-in-Assessments_Sep26_v5.pdf.pdf
[2] International Baccalaureate. Statement from the IB about ChatGPT and artificial intelligence in assessment and education. https://ibo.org/news/news-about-the-ib/statement-from-the-ib-about-chatgpt-and-artificial-intelligence-in-assessment-and-education/ (accessed 16 November 2026; no date shown on the page)
[3] Ofqual (11 December 2025). Malpractice in GCSE, AS and A level: summer 2025 exam series. Ofqual/25/7282/1. 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
[4] College Board. Guidance for Artificial Intelligence Tools and Other Services [section "AI Guidance for the 2026-27 Academic Year"]. AP Central. https://apcentral.collegeboard.org/exam-administration-ordering-scores/administering-exams/exam-policies/artificial-intelligence-tools (accessed 16 November 2026)
[5] Ofqual (16 July 2026). Ofqual's approach to regulating the use of artificial intelligence in the qualifications sector. Ofqual/26/7350. GOV.UK (first published 24 April 2024). https://www.gov.uk/government/publications/ofquals-approach-to-regulating-the-use-of-artificial-intelligence-in-the-qualifications-sector/ofquals-approach-to-regulating-the-use-of-artificial-intelligence-in-the-qualifications-sector--2
[6] College Board (6 October 2025). New Research: Majority of High School Students Use Generative AI for Schoolwork. College Board Newsroom. https://newsroom.collegeboard.org/new-research-majority-high-school-students-use-generative-ai-schoolwork
[7] Joint Council for Qualifications. The JCQ AI Use in Assessments: Your role in protecting the integrity of qualifications [web page]. https://www.jcq.org.uk/exams-office/malpractice/artificial-intelligence/ (accessed 16 November 2026)
[8] Joint Council for Qualifications. Malpractice [web page]. https://www.jcq.org.uk/exams-office/malpractice/ (accessed 16 November 2026)
[9] Ofqual (11 December 2025). Background information for malpractice in GCSE, AS and A level: summer 2025 exam series. Ofqual/25/7282/2. GOV.UK. https://www.gov.uk/government/statistics/malpractice-in-gcse-as-and-a-level-summer-2025-exam-series/background-information-for-malpractice-in-gcse-as-and-a-level-summer-2025-exam-series
[10] Joint Council for Qualifications. Preventing AI Misuse in Assessments: A summary for teachers. https://www.jcq.org.uk/wp-content/uploads/sites/2/2026/09/JCQ-Preventing-AI-Misuse-in-Assessments-A-summary-for-teachers-effective-from-Sept-26.pdf (accessed 16 November 2026)
[11] Ofqual. Malpractice in GCSE, AS and A level: summer 2026 exam series [official statistics announcement, release date 10 December 2026, confirmed]. GOV.UK. https://www.gov.uk/government/statistics/announcements/malpractice-in-gcse-as-and-a-level-summer-2026-exam-series (accessed 16 November 2026; page last updated 9 November 2026)
How to cite
IAPQB (2026). When AI helps with coursework: how exam boards set rules for declaring AI use and for checking that coursework is the student's own. IAPQB Report No. 13. International AI Practice Qualification Board. https://iapqb.org/research/013-ai-in-coursework-rules.html
About this report
Prepared by IAPQB. Each fact comes from the sources listed, and each source was opened and checked on 16 November 2026; anything that could not be checked was left out. AI tools were used to search and to draft, and IAPQB checked every claim against its source and is responsible for the text. IAPQB runs an AI work qualification; this report does not promote it. Corrections: contact@iapqb.org or an issue at github.com/iapqb/iapqb.github.io. Licence: CC BY 4.0.
Archived with the series on Zenodo: doi.org/10.5281/zenodo.24248199 · Markdown sources: github.com/iapqb/research