SAGE Assessment Policy for AI-Integrated Assessment

Student-facing policy

SAGE Assessment Policy (SAGE-AP)

This resource is based on the published paper:
Mahmoud Elkhodr and Ergun Gide, “From Permission to Pedagogy: The Structured AI-Guided Education Assessment Policy (SAGE-AP) for Generative AI in Higher Education,” Education Sciences, vol. 16, no. 6, art. 986, 2026. doi: 10.3390/educsci16060986

Policy for student use of generative AI in assessment

Where generative AI is permitted, students must use it in a way that supports learning, disciplinary judgement, and transparent authorship. AI may assist the work, but it must not replace your own analysis, interpretation, verification, or accountability.

Version 3.1 Effective date: 18 June 2026 Elkhodr & Gide, 2026 CC BY 4.0

1. Scope and unit-level precedence

This policy applies to assessments in which students are required, encouraged, or permitted to use AI in their studies. It may be adopted with or without the SAGE label. When used independently, its full substantive expectations remain unchanged: student authorship, verification, disclosure, responsible refinement, reflection, and defensible work. This policy should be read together with the relevant institutional academic integrity policy, privacy requirements, and any assessment-specific instructions issued by the teaching team.

Always refer to your course outline or unit profile for the specific rules governing AI use in each assessment. Each course and each assessment task may have different guidelines regarding AI tools. It is your responsibility to check whether AI use is permitted and to what extent before beginning any task. The rules for one assessment do not automatically apply to another, even within the same unit.

2. You are the author of your work

The purpose of assessment is to demonstrate what you know, what you can do, and how you reason within your discipline. AI is permitted as a support tool, but the final submission must reflect your own understanding, your own analysis, and your own judgement.

Why this matters: your degree certifies that you have developed specific capabilities. If AI performs the intellectual work and you simply present the result, the assessment cannot verify your learning.

What this means in practice:

In essays, reports, and written assessments, your own argument, synthesis, and critical evaluation must be visible.
In coding and technical tasks, you must understand, explain, and be able to modify the code you submit.
In presentations, the substantive argument and spoken explanation must be your own.
In lab reports and practical work, your own interpretation of data and clinical or design reasoning must drive the submission.
In research tasks, the identification, reading, evaluation, and synthesis of sources must be performed by you.
In group assessments, your individual intellectual contribution must remain distinguishable.
In reflective tasks, the reflective content must be genuinely yours. AI cannot reflect for you.
In creative and design work, AI-generated outputs must be disclosed and must not be presented as your original creative output unless the assessment explicitly permits this.

The test: if you cannot explain why a particular point is in your submission, what source supports it, or how you arrived at a conclusion, the work does not satisfy the authorship requirement — regardless of how much you edited the AI output.

3. What you may use AI for

Subject to the specific conditions in your assessment instructions, the following uses of AI are generally permitted. In each case, the intellectual responsibility remains with you.

Starting and structuring: using AI to generate an initial draft, outline, list, or structural suggestion after you have first established your own understanding of the task (for example, by writing notes, an outline, or a preliminary attempt). In the SAGE framework, this corresponds to the Generate step.
Checking and verifying: using AI to test a response against task requirements, case context, or disciplinary standards, provided you then independently verify the result against authoritative sources. In SAGE, this corresponds to the Evaluate step.
Receiving and judging critique: asking AI to act as a reviewer or critic of your reasoning, then independently deciding which feedback to accept and which to reject. In SAGE, this corresponds to the AI Critic step.
Refining expression: using AI to check clarity, grammar, structure, or expression after the substantive content has been completed by you. An acknowledgement statement is still required.
Search and discovery: using AI to generate search keywords, suggest structural options, or produce plain-language summaries of sources you have already verified, provided that all sources are independently found, read, and confirmed by you through library databases or other authoritative channels.
Reflection support: using AI to check whether a reflective piece addresses all required dimensions, without using AI to generate the reflective content itself.
Code assistance: using AI to explain error messages, suggest debugging approaches, or generate small code snippets that you then understand, modify, test, and can explain in full.
Visual and design support: using AI to generate initial concept sketches, diagram layouts, or design variations that you then substantially develop, refine, and justify using discipline-specific principles, with the AI starting point disclosed.

The key condition: in every case, the student must remain the decision-maker. AI produces raw material; you produce the judgement, the verification, and the final reasoning.

4. What you must not use AI for

The following uses of AI are not permitted because they transfer the intellectual work from the student to the tool, making it impossible for the assessment to verify your learning.

Submitting AI-generated work as your own: submitting AI-generated text, analysis, interpretation, code, designs, or conclusions as if they were your own original work. This applies even if you edited the output afterwards, and even if AI use is correctly attributed, where the assessment required your own intellectual work.
Delegating core assessment tasks: asking AI to write the final answer, essay, discussion, literature review, reflection, care plan, policy analysis, code explanation, lab interpretation, design rationale, or any other component that the assessment was designed for you to produce.
Using unverified AI-generated references: citing references, quotations, statistics, case law, clinical guidelines, technical standards, or factual claims suggested by AI without independently locating and verifying the original source. AI frequently fabricates plausible-sounding references that do not exist.
Outsourcing evaluative judgement: allowing AI to make the analytical, interpretive, or evaluative decisions that the assessment requires you to make, without applying your own independent reasoning and verification.
Fabricating process evidence: creating, back-filling, or simulating process logs, prompt trails, decision records, or evidence of engagement that do not reflect the work actually undertaken.
Generating false reflection: using AI to produce reflective content that purports to describe your own learning experience, self-awareness, or professional development. Reflection must be genuinely personal.
Concealing AI use in group work: using AI to produce your contribution to a group assessment without disclosing this to the group or in the process log, effectively outsourcing your individual role.
Entering restricted information: entering personal, confidential, proprietary, patient, client, or commercially sensitive information into public AI tools in connection with the assessment (see Clause 8 for detail).
Copyright infringement: entering copyrighted material into AI tools without authorisation, or submitting AI-generated content that reproduces a substantial part of an existing copyrighted work (see Clause 9 for detail).
Producing work you cannot defend: submitting work produced with AI assistance that you are unable to explain, justify, or reproduce under supervised conditions. If you cannot defend it, you do not own it.

5. Verify everything AI tells you

AI generates plausible-sounding text by predicting what words are likely to follow other words. This means it can produce content that is factually wrong, fabricated, outdated, or misleading. You must independently verify every important claim before relying on it.

Why this matters:

Hallucination: AI may cite sources that do not exist, fabricate quotations, or invent statistics.
Training cutoffs: AI may not reflect recent developments, legislative changes, or updated guidelines.
Low-quality sources: AI does not distinguish between blog posts and peer-reviewed literature.

What to do: do not cite any AI-suggested reference unless you have personally found, read, and verified it. Check claims against the authoritative sources specified for your assessment. Consult your teaching team or library if unsure.

6. Be aware that AI output may be biased

AI tools are not neutral. They can produce outputs reflecting systematic biases:

Cultural bias: training data is predominantly Western and English-language. Outputs may not apply to Australian, Asia-Pacific, Indigenous, or other contexts.
Confirmation bias: AI tends to reinforce mainstream views and underrepresent minority or contested perspectives.
Authority bias: AI may overweight frequently cited sources regardless of quality or currency.

You are responsible for assessing whether AI output is appropriate for the disciplinary, cultural, and contextual requirements of your task.

7. Disclose and attribute AI use correctly

All AI use must be disclosed as specified in the assessment instructions:

Quoting or paraphrasing AI content: in-text citation or footnote (APA, IEEE, Harvard, etc.).
AI used for editing or translation: acknowledgement statement at the end of the document.
Detailed process evidence required: complete the structured declaration as specified.

Important: correct attribution does not automatically resolve an integrity issue. A predominantly AI-generated submission remains problematic even if properly acknowledged, because the intellectual work has been outsourced.

8. Protect personal information and confidential data

Do not enter personal, confidential, or restricted information into public AI tools. This includes:

names, addresses, emails, phone numbers, student IDs (yours or others’);
thesis work, unpublished research, proprietary code, institutional documents;
patient, client, or case data; commercial-in-confidence material.

What to do instead: use institutionally approved AI platforms where available — these typically do not retain your data or use it for training. Check with your institution which tools are approved. Do not use AI tools to detect AI-generated content in your own or others’ work.

9. Respect copyright and intellectual property

What you enter: pasting copyrighted material (textbook passages, articles, theses, code) into AI may constitute infringement without the copyright holder’s permission.

What AI produces: outputs may reproduce existing copyrighted works. You are responsible for ensuring that nothing in your submission infringes copyright on either side.

10. Understand what counts as AI use

This policy applies to deliberate, task-directed use of generative AI. Standard spell-checking, predictive text, and basic autocomplete do not count unless the assessment says otherwise. However, tools like Microsoft Copilot in Word, AI paraphrasing tools, or AI code generation assistants do fall within scope. If unsure, ask the teaching team before proceeding.

11. These rules apply to all forms of AI output

This policy covers text, code, images, diagrams, audio, video, data analysis, and design prototypes. If AI produced or substantially shaped any artefact in your submission, the same obligations apply.

12. How the SAGE steps work

Generate: use AI to produce an initial output after establishing your own understanding. Why: starting from your own thinking ensures AI supports learning rather than bypassing it.
Evaluate: compare AI output against authoritative sources. Assess for accuracy, source quality, and bias. Why: AI generates plausible content, not necessarily correct content.
Refine: decide what to accept, modify, or reject, with reasons. Why: your decisions are the core evidence of learning.
AI Critic: ask AI to critique your work, then judge the critique independently. Why: the AI does not get the final word; you do.
Reflect: identify where AI helped, where it failed, and what this reveals about AI in your discipline. Why: this builds the critical AI literacy that will serve your career.
Defend: demonstrate under supervision that you understand and can justify your reasoning. Why: a polished submission alone cannot prove learning occurred.

13. Academic integrity

The following are serious integrity matters:

submitting predominantly AI-generated work, even if attributed, where the assessment required your own intellectual work;
fabricating or back-filling process logs after the fact;
using AI secretly in a group task contrary to the group agreement;
being unable to explain or defend work submitted under your name.

This policy links to the institution’s broader academic integrity procedures.

14. What happens if you cannot fully defend your work

Competence matter: if you engaged in good faith but demonstrate gaps in understanding during Defend, this may result in feedback, resubmission, or grade adjustment — not necessarily a misconduct finding.

Misconduct matter: fabrication, concealment, or deliberate substitution is handled through the institution’s integrity procedures.

15. Student responsibility statement

By submitting work under this assessment, you confirm:

the final submission reflects your own understanding, analysis, and reasoning;
any AI use complied with the stated conditions of this assessment and your course outline;
all factual claims and sources were independently verified by you;
all material AI use has been disclosed as required;
you have not entered personal or restricted information into public AI tools;
you are able to explain and defend the reasoning in the work.

Recommended short declaration for assessment briefs.

This assessment permits generative AI use only within the SAGE process. AI may assist the work, but it must not replace your own reasoning, source verification, or authorship. You must follow the six SAGE steps (Generate, Evaluate, Refine, AI Critic, Reflect, Defend), retain any required records of AI use, disclose AI involvement as specified, and be prepared to defend your work under supervised conditions. Always check your course outline for the specific AI conditions applying to this task.

Examples

What responsible and irresponsible AI use looks like

Acceptable use

Examples that support learning

Essay: write an outline first, then ask AI for feedback on gaps without asking it to rewrite.
Coding: write a function, use AI to explain an error message, then fix the code yourself.
Presentation: use AI to suggest slide sequence; write all analysis and speaker notes independently.
Lab report: complete the experiment, record data, ask AI to explain a statistical concept. Interpretation remains yours.
Research: use AI to generate search keywords, then independently locate, read, and synthesise sources.
Group project: team uses AI to brainstorm structure; each member writes their own section and documents AI use.
Reflective journal: write your own reflection, then use AI to check coverage of required dimensions.
Design portfolio: use AI for initial concept sketches, then substantially redesign and disclose the AI starting point.

Not acceptable use

Examples that undermine integrity

Essay: ask AI to write the full report and submit with cosmetic changes, even if acknowledged.
Coding: generate an entire codebase and cannot explain how it works.
Presentation: AI generates the spoken argument; student memorises and presents it as their own.
Lab report: enter raw data into AI and submit the AI interpretation as your own analysis.
Research: copy AI-generated references without independently verifying the sources exist.
Group project: use AI to generate your section without telling the group.
Reflective journal: ask AI to write a reflection on experiences you did not describe.
Design portfolio: submit AI-generated design assets as your original work without disclosure.
Any assessment: fabricate a process log after the fact to appear compliant.

Worked example: acceptable

Health sciences

A student drafts a paragraph interpreting a verified journal article, then asks AI:

I wrote the following paragraph. Here is the article excerpt. Identify any potential misinterpretations. Do not rewrite my paragraph.

Why acceptable: the student produced the paragraph, named the source, and limited AI to feedback. Intellectual labour is visible and attributable.

Worked example: not acceptable

Cybersecurity

A student asks AI:

Write my 1,500-word literature review on cyber risk governance with five references from 2023 onwards.

Why not acceptable: the prompt delegates framing, source selection, synthesis, and writing to AI. Even if edited afterwards, the original reasoning was outsourced.

Worked example: acceptable

Business management

A student in a strategic management unit writes a SWOT analysis of a case company based on the supplied case study materials, then asks AI:

Review my SWOT analysis below against the case facts provided. Identify any claims I have made that are not supported by the case data. Do not rewrite my analysis.

Why acceptable: the student produced the analysis, grounded it in the case materials, and limited AI to a verification role. The strategic reasoning is the student’s own. The AI was used to check the work against evidence, not to generate the analytical judgement.

LMS reminder (copy-paste ready)

AI reminder. Use AI to support your thinking, not to replace it. Start with your own understanding. Check AI output against authoritative sources. Document your decisions. Reflect on what AI got wrong. Be prepared to defend your work. Check your course outline for the specific AI rules applying to this assessment.

AI use declaration

Declaration templates

Three variants are provided below, scaled to the stakes and complexity of the assessment. Educators should select or adapt the version that best fits their context. Not every assessment requires the full evidence log.

Short declaration — for formative tasks, first-year work, or routine assessments

AI Use Declaration (Short) Tool(s) used: Purpose of use: Authorship confirmation: I confirm that the final submitted work reflects my own reasoning and judgement. Any AI use has been disclosed above. I accept responsibility for the accuracy and integrity of all content submitted under my name. Student name: Date:

Standard declaration — for most written and practical assessments

AI Use Declaration (Standard) Tool(s) used: Date(s) of use: Sections or task stages affected: Purpose of use (select all that apply): ☐ Clarification of instructions ☐ Brainstorming or outline support ☐ Language or grammar refinement ☐ Source-search keyword generation ☐ Checking alignment against named sources ☐ Code generation or debugging ☐ Image, diagram, or design generation ☐ Other: _______________ What I verified independently and against which sources: Authorship confirmation: I confirm that the final submitted work reflects my own judgement, interpretation, and reasoning. I have not submitted AI-generated content as my original work. All material AI use has been disclosed. I accept responsibility for the accuracy, integrity, and disclosure of all content submitted under my name. Student name: Student number: Date:

Full evidence log — for capstone, defended, or research-intensive assessments

AI Use Declaration (Full Evidence Log) Tool(s) used: Date(s) of use: Sections or task stages affected: Purpose of use (select all that apply): ☐ Clarification of task instructions ☐ Brainstorming or outline support ☐ Language, grammar, or expression refinement ☐ Source-search keyword generation ☐ Checking alignment of my writing against named sources ☐ Generating an initial draft after producing a human baseline ☐ AI Critic step: AI assigned as reviewer or critic ☐ Code generation, debugging, or explanation ☐ Image, diagram, or design generation ☐ Data analysis support ☐ Translation or multilingual support ☐ Other (specify): _______________ What I provided before using AI (human baseline): What I verified independently and against which sources: Summary of Refine decisions: ☐ Accepted (with justification) ☐ Modified (with justification) ☐ Rejected (with justification) How I handled the AI Critic step: Reflection on limitations, biases, or errors identified: Authorship confirmation: I confirm that the final submitted work reflects my own judgement, interpretation, and reasoning. I have not submitted AI-generated text, analysis, code, images, or conclusions as my original work. I have disclosed all material AI use as required. I accept responsibility for the accuracy, integrity, and disclosure of all content submitted under my name. Student name: Student number: Date:

Adaptation note: educators should modify the full log to match the assessment type. For coding tasks, replace the Refine summary with a code change log. For design portfolios, include a visual decision record.

FAQ

Common questions about AI use in assessment

Can AI write part of the assignment if I edit it heavily afterwards?

Editing AI-written text does not automatically restore authorship if the original reasoning was outsourced. The question is whether the intellectual labour behind the submission is genuinely yours, not whether the final text was typed by you.

What if AI gives a very good answer immediately?

A convincing answer does not remove the requirement to verify claims, identify potential errors, and be able to defend the reasoning. A fluent AI output is a starting point, not a finished product.

Can I use AI just for grammar and expression?

Usually yes, where permitted. However, grammar support must not drift into content generation or rewriting that conceals the source of reasoning. An acknowledgement statement is typically required even for editing use.

Do I need to save my prompts and outputs?

Yes, unless your lecturer explicitly states otherwise. These records are part of the evidence of engagement and authorship. Even when not submitted, they may be requested.

What happens in the Defend step?

You may be asked to explain reasoning, justify choices, respond to challenge questions, walk through code, or reproduce thinking under supervised conditions. The format and duration will be specified in the assessment instructions.

What if I try my best but cannot fully defend the work?

An inadequate defence does not automatically mean misconduct. If you engaged in good faith, this is treated as a learning matter. Wilful misconduct (fabrication, concealment) is a separate and more serious matter.

Does AI have biases I should know about?

Yes. AI can reflect Western-centric, confirmation, and authority biases, including the reinforcement of dominant views and the overweighting of popular sources. You are responsible for assessing whether output is appropriate for your specific task and context.

Can I paste copyrighted material into AI?

Not without the copyright holder’s permission. Also be aware that AI outputs may reproduce copyrighted material. You are responsible for both sides.

Does predictive text or autocomplete count as AI use?

The policy applies to deliberate, task-directed use of generative AI tools, not to incidental spell-checking or predictive text. However, tools like Microsoft Copilot in Word or AI code assistants that produce substantial content do count. If unsure, ask before proceeding.

How does this work in a group assessment?

Each member’s individual contribution should remain distinguishable. The group should agree how AI will be used, document this, and attribute AI-assisted elements to specific members.

I cited AI correctly. Is that enough?

Not necessarily. Correct attribution is necessary but not sufficient. A predominantly AI-generated submission is problematic even if acknowledged, because the intellectual work was outsourced.

Framework-neutral alternative

Principles-based policy for responsible AI use

This is the framework-neutral variant of the policy, stated as fourteen principles in plain language for educators who wish to adopt it without naming SAGE. It carries the same expectations as the full policy and may be adapted under CC BY 4.0 with attribution.

Policy for responsible use of generative AI in assessment

Generative AI may be used in this assessment only as a support for learning and drafting. It must not replace your own reasoning, analysis, interpretation, or responsibility for the final submission.

Principle 1: Check what is permitted. Confirm whether and how much AI use is allowed for this specific task before you begin. Rules for one assessment do not carry over to another.

Principle 2: You are the author. The final submission must reflect your own understanding, analysis, and judgement. If you cannot explain why a point is included, what supports it, and how you reached it, the work does not meet this requirement.

Principle 3: Permitted uses. Where permitted, you may use AI to start, structure, check, critique, and refine, provided verification and final judgement remain with you.

Principle 4: Prohibited uses. Do not submit AI-generated work as your own, delegate the core task, cite unverified AI references, outsource judgement, fabricate process evidence, or conceal AI use.

Principle 5: Verify everything. Check every important AI-generated claim, reference, or statistic against authoritative sources. Do not cite anything you have not personally checked.

Principle 6: Watch for bias. Assess AI output for cultural, confirmation, and authority bias. Do not assume it is neutral, current, balanced, or appropriate to your context.

Principle 7: Disclose and attribute. Disclose AI use in the required form. Disclosure alone does not make a predominantly AI-generated submission acceptable.

Principle 8: Protect data. Do not enter personal, confidential, or restricted information into public AI tools. Use approved platforms where available.

Principle 9: Respect copyright. Do not enter copyrighted material into AI tools without permission, and do not submit output that reproduces protected work.

Principle 10: Know what counts. Deliberate, task-directed AI use is in scope, including AI paraphrasing and code generation. Incidental spell-checking and autocomplete are not, unless the assessment says otherwise.

Principle 11: All output types. These rules apply equally to AI-generated or AI-shaped text, code, images, audio, data analysis, and design.

Principle 12: Work and document. Work from your own understanding, then check AI output, record what you accepted, modified, or rejected and why, and note where AI helped and where it failed.

Principle 13: Defensible authorship. Be ready to explain, justify, or demonstrate the reasoning in your submission under supervised conditions at any time. If you cannot defend it, you do not own it.

Principle 14: Integrity and good-faith gaps. Predominantly AI-generated work, fabricated evidence, concealed AI use, and inability to explain your work are integrity matters. A good-faith gap in understanding is treated as a competence matter, handled differently from misconduct.

Student declaration. By submitting this work, I confirm that it reflects my own reasoning, complies with the stated AI conditions, has been independently verified, discloses any material AI use, contains no restricted information entered into public AI tools, and can be explained and defended by me.