Generative AI has not made assessment impossible; it has made weak assessment easier to expose. If a task can be completed convincingly by producing generic text with little subject understanding, the assessment may have been measuring product submission more than learning. Schools now need systems that value process, application, explanation, judgement and evidence of student thinking.
The answer is not a blanket return to handwritten tests or a blanket acceptance of AI. Authentic assessment asks students to use knowledge in meaningful contexts and makes learning visible through more than one piece of evidence. A school-wide design should also define when AI is prohibited, when it is permitted, how its use is acknowledged and what students remain responsible for knowing and doing independently.
What makes an assessment authentic?
Authentic assessment requires students to apply knowledge or skills in a context that resembles the thinking of the discipline or a meaningful real-world problem. It might involve investigation, design, analysis, performance, explanation, critique or creation for an audience.
Authenticity is not the same as elaborate projects. A short source evaluation or oral explanation can be highly authentic if it reveals disciplinary judgement.
Why has generative AI changed assessment validity?
AI can produce fluent summaries, essays, code, images and explanations quickly. Tasks that award most credit for generic polished output may no longer provide strong evidence of what the student can do independently. This creates an assessment-validity problem before it becomes a disciplinary problem.
Redesign should therefore ask what evidence would convince a teacher that the student understands the learning, regardless of the tool used.
Should schools ban AI from all assessment?
A universal ban is difficult to enforce and educationally limiting. Some assessments should remain clearly AI-free because independent knowledge or skill is the target. Other tasks can permit AI as a tool while assessing judgement, verification, refinement or reflection.
The important element is explicit task-level guidance. Students should not have to guess whether a particular use is acceptable.
How can process evidence strengthen assessment?
Collect planning, drafts, annotations, data, design choices, checkpoints and reflections. Process evidence makes learning visible over time and allows teachers to give feedback before final submission. It also makes it harder for a single generated product to substitute for genuine engagement.
Do not require excessive documentation merely to police AI. Select a few meaningful checkpoints that already support learning.
What is the role of oral defence or viva-style questioning?
A brief oral conversation can test ownership. Ask the student to explain a claim, justify a source, reconstruct a method, respond to a counterargument or describe what they would change. The purpose is not interrogation; it is additional evidence of understanding.
Oral defence can be sampled for larger classes rather than attached to every task. Teachers need clear criteria and equitable procedures.
How can in-class evidence complement take-home work?
Use short supervised tasks before or after extended work. A student might analyse a related source, write a paragraph, solve a parallel problem or explain a concept. The comparison helps the teacher judge whether the submitted product aligns with demonstrated capability.
This approach is more educationally useful than relying on imperfect AI-detection tools.
How should AI use be acknowledged?
Schools can require students to disclose tools used, the purpose of use and significant prompts or outputs where relevant. Expectations should match age and task. The goal is transparency and attribution, not bureaucratic prompt logging.
Teach students that responsibility remains with the author. AI output can be inaccurate, biased or inappropriate; verification is part of academic integrity.
What kinds of tasks are more resilient to generic AI output?
Tasks anchored in local data, classroom experiments, personal observations, specific texts, unique datasets, iterative design or live collaboration are harder to complete meaningfully through generic generation. Ask students to make decisions and defend them.
However, do not design artificial tasks solely to “beat AI.” The assessment should still represent worthwhile learning.
How can teachers assess critical use of AI?
Where AI is permitted, assess the student’s ability to evaluate output. Students can compare generated responses with disciplinary sources, identify errors, improve reasoning, test assumptions or explain why they rejected a suggestion.
This shifts the task from content production to informed judgement, while still requiring foundational knowledge.
How should rubrics change?
Rubrics may need greater emphasis on reasoning, evidence, application, process, reflection and oral explanation rather than polish alone. Criteria should describe quality of thinking that can be observed.
Avoid vague “originality” criteria that are difficult to apply fairly. Define what students are expected to demonstrate.
What should remain assessed without AI?
Students still need independent fluency in foundational knowledge and skills. Schools should identify moments when memory, calculation, reading, writing, practical technique or disciplinary reasoning must be demonstrated unaided. These become part of a balanced assessment system.
AI literacy depends on knowledge; students cannot evaluate output well if they know too little about the subject.
How can schools maintain fairness when access to AI varies?
If AI is required for a task, the school should provide equitable access or an equivalent route. Do not assume all students have the same devices, subscriptions, connectivity or home support. Privacy restrictions may also limit use for younger students.
Assessment design should not reward purchasing power.
How should teachers respond to suspected misuse?
Follow a clear process based on evidence and conversation. AI detectors can be unreliable and should not be treated as definitive proof. Compare drafts, in-class performance, references and the student’s ability to explain the work.
Consequences should align with published academic-integrity policy and distinguish deliberate deception from misunderstanding of new rules.
What professional development do teachers need?
Teachers need time to test generative tools, examine assessment vulnerabilities, redesign tasks, calibrate expectations and discuss real cases. A policy written without teacher capability will be applied inconsistently.
Departments should decide what authentic evidence looks like in their disciplines; a good science assessment is not identical to a good literature or arts assessment.
How should parents understand the new assessment model?
Explain why familiar homework products may carry less evidential weight and why process, discussion or supervised application matters. Clarify acceptable AI use and the school’s commitment to both innovation and independent learning.
Parents can support integrity more effectively when rules are understandable rather than framed as a technology panic.
How can leaders audit assessment quality school-wide?
Sample tasks across subjects and ask: What learning is being measured? Could a student submit a plausible product without understanding? What process evidence exists? Is tool use explicit? Is the task accessible and appropriately challenging?
Use the audit to identify design patterns for professional learning rather than blaming individual teachers.
What does a balanced AI-era assessment system look like?
It combines supervised and unsupervised evidence, product and process, individual and collaborative work, independent fluency and responsible tool use. No single format carries the entire judgement.
Generative AI should push schools toward better assessment, not simply more surveillance. When tasks require students to know, apply, explain and defend their thinking, assessment becomes more authentic and more educationally valuable whether AI is present or not.
Extending the discussion: a school leader’s perspective
Technology should solve an educational problem
The first question about any digital tool should not be whether it is innovative. It should be what educational problem it solves. A platform may save teacher time, improve feedback, make practice more adaptive or widen access to resources. If the problem is unclear, technology can easily become expensive decoration.
AI literacy is now part of literacy
Students increasingly encounter generative AI in search, writing, coding, revision and creative work. Schools therefore need to teach more than rules about whether AI is allowed. Learners should understand that AI can produce confident errors, reproduce bias, blur authorship and create privacy risks.
Digital dharma: freedom with responsibility
A useful Indic lens for technology is dharma understood as responsible action. Digital spaces give young people extraordinary freedom, but freedom without responsibility can harm others and damage the self.
Protect privacy by design
Student data protection should not depend only on individual caution. Schools need institutional safeguards around accounts, permissions, app approval, data retention, parental consent and vendor access.
Teach verification, not just search
The abundance of information has changed the meaning of research. Students can find an answer quickly, but the harder skill is deciding whether it deserves trust.
Balance screen efficiency with human learning
Not every learning experience improves when moved to a screen. Discussion, handwriting, practical work, movement, reading from physical texts and face-to-face collaboration still have important educational value.
Teacher capability matters more than software features
A powerful platform in the hands of an unsupported teacher will often be used superficially. Adoption depends on professional learning, time to practise, clear expectations and access to help when something goes wrong.
Create age-appropriate digital progression
Digital citizenship should develop progressively. Younger students need simple habits around privacy, kindness, screen balance and asking an adult for help. Older students can engage with digital identity, algorithms, misinformation, intellectual property, AI, online relationships and data ethics.
Involve parents without creating panic
Parents often receive digital-safety messages only after a problem occurs. Schools can be more effective by building regular, calm communication around emerging risks and practical family habits.
Audit the impact, not just the usage
High login numbers do not necessarily mean a digital tool is improving learning. Schools should look for evidence such as reduced teacher workload, improved feedback cycles, better access, stronger student practice or clearer progress information.
Continue exploring related ideas
For a wider perspective, you may also find these related articles useful: How Should International Schools Design Hybrid Learning Infrastructure for Future-Ready Education?, Digital Citizenship in Schools and What AI Literacy and Guardrails Do School Leaders Need for Faculty and Students?. For the wider policy context, see EdTech & AI in Schools.