For decades, teachers could assign an essay, research paper or take-home problem set and assume that the finished work mostly reflected what a student could do.
Artificial intelligence has changed that assumption.
A student can now ask an AI chatbot to draft an essay, solve mathematics problems, write computer code, summarize a book or improve an assignment in seconds.
That does not mean every student using AI is cheating.
But it creates a difficult question for schools:
If AI can produce the assignment, how can teachers know what the student actually learned?
In 2026, schools and universities across the United States, United Kingdom, Canada and Australia are responding by redesigning exams and homework.
The future of assessment may involve fewer traditional take-home essays and more in-class work, oral examinations, presentations, personalized questions and assignments that document the student’s thinking process.
1. Traditional Homework Has Become Harder to Verify
The central problem is simple.
Outside a supervised classroom, teachers cannot always know how much of an assignment was completed by the student.
The University of Waterloo in Canada now tells instructors that generative AI can potentially complete nearly any unsupervised assessment and recommends redesigning tasks that AI performs particularly well, including fact-based quizzes and generic online discussion posts.
That does not mean homework is disappearing.
Instead, its purpose may change.
Homework can become practice, while supervised exams, presentations or conversations verify that students actually understand the material.
2. England Is Tightening Rules Around AI-Generated Coursework
England is taking the issue particularly seriously.
In March 2026, Ofqual—the regulator responsible for qualifications including GCSEs and A levels—released new resources specifically addressing AI and coursework.
The rule is clear: students must not use AI to generate assessed work and then present that material as their own. AI use must be acknowledged where required.
Ofqual says penalties for assessment malpractice can be severe.
Across all forms of malpractice in 2025, 1,125 cases resulted in students losing an entire GCSE or A-level qualification, while nearly 2,000 cases resulted in marks being deducted.
Ofqual has also warned that coursework completed outside controlled conditions is particularly vulnerable to AI misuse, while supervised examinations are much less exposed because students cannot normally access AI tools during them.
3. The Essay May No Longer Be Enough
For centuries, essays have been one of education’s most common ways of measuring knowledge.
But generative AI can now produce a polished essay extremely quickly.
EDUCAUSE, which studies technology in higher education, argues that AI has exposed a fundamental weakness in assessment: a finished essay is an artifact of learning, but it does not always prove the student possesses the understanding reflected in that artifact.
That is pushing institutions toward what educators call authentic or process-based assessment.
Instead of simply asking students for a 2,000-word essay, a teacher might require:
- the initial research question
- an outline
- source notes
- early drafts
- revisions
- an explanation of decisions
- a final presentation
- a short oral defence
The teacher can then see how the student’s ideas developed.
The 2026 EDUCAUSE Horizon Report says higher education is already moving toward authentic, process-based demonstrations of learning as AI makes conventional assessments increasingly difficult to trust.
4. Oral Exams Are Making a Comeback
One of the oldest forms of assessment could become one of the most useful in the AI era.
The viva voce, or oral examination, requires students to answer questions and explain their reasoning in real time.
At the University of Toronto, one Winter 2026 course replaced its traditional written final examination with a 20-minute one-on-one oral conversation.
Students were given possible questions beforehand, but the actual questions were randomly selected during the examination. No devices were allowed, meaning students had to demonstrate genuine understanding in conversation.
5. Some Homework Is Moving Back Into the Classroom
Another emerging strategy is turning take-home assignments into supervised activities.
A first-year chemistry course at the University of Toronto redesigned its laboratory assessment in Winter 2026.
Instead of students completing graded lab reports at home, calculations, discussions and conclusions are now completed during the laboratory session itself.
Students can still use AI-supported resources for some preparation, but the assessed analytical work happens in person.
This illustrates the new philosophy:
AI can help students prepare—but schools still need moments when students demonstrate skills independently.
6. Canada Is Making AI Rules More Explicit
Universities are also trying to eliminate another major problem: students simply not knowing whether AI is allowed.
At the University of British Columbia’s Vancouver campus, a new policy effective September 1, 2026 requires course syllabi to include statements explaining academic-integrity expectations and how generative AI may—or may not—be used in submitted work.
The goal is important because AI rules can differ enormously between courses.
A coding professor may encourage students to use AI.
A writing instructor may allow it only for brainstorming.
Another course may prohibit it during major assessments.
Instead of vague institution-wide bans, schools increasingly need assessment-by-assessment instructions.
7. Australia Is Redesigning Assessment Instead of Relying Only on AI Detection
Australia’s higher-education regulator, TEQSA, is also urging universities to redesign assessment for a world where generative AI is widely available.
In June 2026, TEQSA published new guidance on maintaining reliable evidence of student learning in an AI-integrated environment. It emphasizes abilities such as critical thinking, evaluative judgement and ethical reasoning.
Australian guidance also promotes authentic assessments that require personal engagement and real-world problem solving, making it harder for students simply to outsource the entire assignment to AI.
This is an important shift.
Schools initially focused heavily on detecting AI-generated writing.
The emerging approach is different:
Design an assignment that AI cannot complete successfully without meaningful student involvement.
8. Some Exams May Actually Allow AI
Schools are not moving toward a future where every assessment bans artificial intelligence.
Some assignments may intentionally require students to use it.
For example, students could be asked to:
- generate an answer using AI,
- identify errors or weaknesses,
- verify the claims using reliable sources,
- improve the response,
- explain what they changed and why.
That tests a skill students are likely to need in AI-enabled workplaces.
University of Waterloo guidance recommends pairing AI-integrated assignments with human-centered components and transparent rules about when AI is expected, permitted or prohibited.
So the classroom of the future could have two types of assessment:
AI-free assessments that prove independent knowledge.
and
AI-enabled assessments that prove students know how to use technology intelligently.
9. The United States Is Emphasizing Educational Value Over Technology for Its Own Sake
In August 2026, the US Department of Education issued updated guidance on responsible classroom technology.
The department urged states and schools to focus on instructional value, evidence, educator judgement, transparency for parents and measurable student outcomes when adopting new technology. (ed.gov)
That principle applies directly to assessment.
Schools should not redesign exams merely because AI exists.
They need to ask whether the new assessment actually gives teachers better evidence that students learned something.
10. AI May Change What Schools Value
Perhaps the biggest transformation will not be the format of exams.
It will be what exams measure.
When AI can instantly recall facts, summarize information and generate basic text, schools may place greater value on abilities that are harder to automate:
- critical thinking
- reasoning
- source evaluation
- creativity
- communication
- problem solving
- personal judgement
- explaining decisions
A 2026 EDUCAUSE survey of 438 faculty and staff involved in learning assessment found growing pressure to rethink assessment practices and clearer policies around when students should and should not use AI.