Designing Authentic Online Assessments When Students Can Use Generative AI

Generative AI has made it easier to produce fluent responses, but polished text does not necessarily demonstrate genuine learning. Online assessment authenticity now depends on clear rules, thoughtful task design, and proportionate verification, not simply increased surveillance.

This blog by All Assignment Help explores how students can navigate online courses while managing academic pressure. It also examines the importance of maintaining assessment integrity and offers practical strategies to address the challenges posed by AI-powered language models.

Read more: Strategies to Perform Well in Your Online Mock Tests

What authentic assessment means in an AI-enabled environment

An authentic assessment asks students to apply knowledge in a meaningful context and provides evidence that the submitted work reflects the intended learning outcome. Authenticity does not require every task to imitate a workplace. It requires alignment between what the course claims to teach and what the assessment actually measures.

Institutions should define when AI is prohibited, permitted for limited functions or expected as part of the task. A vague warning to “use AI responsibly” leaves students and markers to interpret different boundaries.

Where identity verification is necessary, the related guide explains the student-facing questions around online exam proctoring.

State permitted and prohibited uses in the task itself

Instructions can specify whether students may use AI for brainstorming, language feedback, coding assistance, source discovery or no part of the assessment. They should also explain disclosure and citation requirements. Rules may differ between assignments in the same course, so a programme-wide statement is not always enough.

Give instructors and students separate responsibilities

Instructors are responsible for aligning the task with the learning outcome, explaining the rules and using a fair review process. Students are responsible for reading those rules, preserving evidence when disclosure is required and asking before using a tool in an unclear situation. Institutions support both groups by providing consistent policy, training, accessibility routes and privacy oversight.

This division avoids the unhelpful assumption that integrity is only a student problem or that every instructor must invent an independent AI policy.

Design tasks that reveal the learning process

• Require a proposal, outline or annotated source selection before the final submission.
• Ask students to justify decisions using course-specific material.
• Include drafts, revision notes or a brief process reflection.
• Use local cases, lab observations, datasets or class discussions where appropriate.
• Follow written work with a short oral explanation or demonstration for selected tasks.

No single element proves authorship, but several low-burden checkpoints create a more meaningful record of learning.

Do not treat AI detection as conclusive evidence

Automated detection can produce uncertain results and should not replace a fair academic-integrity process. Decisions should consider the work, the assessment design, documented concerns and the student’s explanation. Institutions need transparent procedures for review and appeal.

Use proctoring only when it is proportionate

Identity verification or remote proctoring may be justified for some high-stakes assessments, but it introduces privacy, accessibility and equity concerns. Institutions should minimise data collection, explain retention, provide accommodations and consider alternative assessment designs.

Teach verification as part of AI literacy

Students need practice checking AI output against original sources, identifying fabricated references, protecting confidential information and documenting assistance. An assessment can ask learners to critique an AI response, correct its reasoning or compare it with scholarly evidence. This measures judgment rather than rewarding uncritical generation.

Markers also need guidance on how disclosed AI assistance affects grading. If permitted use is declared but penalised unpredictably, disclosure becomes less likely. Rubrics should assess the intended knowledge, reasoning and communication rather than the mere presence of a tool.

What students should do when the rules are unclear

Ask the instructor before submission and retain the response. Keep notes about tools used, prompts entered and changes made when disclosure may be required. Do not upload private course materials, personal data or unpublished research to an external system without permission.

Students needing help with preparation should use tutoring, feedback and permitted study support. They can review online class assistance while remaining responsible for work that must demonstrate their own learning.

A practical assessment-authenticity checklist

• The learning outcome is explicit.
• The task requires application, judgment or explanation.
• AI permissions and disclosure rules are stated.
• Students have access to permitted tools and accommodations.
• Process evidence is proportionate to the assessment.
• Suspected misconduct receives human review and due process.
• Personal data collection is necessary, disclosed and protected.

Use disclosure that produces useful evidence

A disclosure requirement should tell students what information to provide and explain why it matters. Asking only “Did you use AI?” reduces a complex process to a yes-or-no answer. A more useful declaration can identify the tool, permitted purpose, stage of work, important prompts or instructions, material retained and checks performed by the student. The level of detail should remain proportionate to the assessment.

Disclosure is not a substitute for assessment design. A long prompt transcript may create administrative work without proving understanding, and it may contain personal or confidential information. Instructors can instead request a short process note, selected draft evidence or an oral explanation of a decision. Students should not be penalised unpredictably for disclosing a use that the instructions permitted.

Policies should also address unequal access. Paid tools, language features and institutionally licensed systems may provide different capabilities. If a particular tool is required, students need comparable access and an alternative where disability, privacy, location or technical constraints prevent its use. Authenticity is strengthened when expectations are clear, achievable and connected to the learning outcome.

Because models and institutional licences change, task instructions should identify the applicable policy version or assessment date. This prevents a later platform update from silently changing the conditions under which earlier student work is judged.

Read more: The Importance of Accreditation in Online Education

Final thought

Assessment authenticity in the ChatGPT era is primarily a design and communication challenge. Clear permissions, meaningful application and fair verification provide stronger evidence of learning than vague rules or automatic suspicion.

Frequently asked questions

Does using ChatGPT automatically constitute cheating?

No universal rule applies. It depends on the assessment instructions and institution’s policy. Undisclosed use is risky when the task requires independent work.

Can an instructor rely on an AI-detection score?

A score should not be treated as proof by itself. Institutions should use fair review procedures and allow the student to respond.

How can online assessments remain authentic without surveillance?

Possible methods include staged submissions, oral follow-ups, personalised application, project work and reflection on decisions. The appropriate design depends on the learning outcome.