A source-based overview of how SAGE is referenced in the official Educator Guide, followed by independent SAGE resources for educators.
CQUniversity’s Guided–Assured Assessment Model Educator Guide presents a practice-oriented framework for assessment in the context of generative artificial intelligence.
The guide describes three assessment-security tiers:
The Educator Guide also includes GenAI Guidance Levels, an assessment-design decision framework, course-level planning considerations, and guidance concerning accessibility, equity and academic integrity.
The Educator Guide incorporates SAGE within its assessment-design material. Page 12 aligns the GenAI Guidance Levels with SAGE Student Orchestration Levels, while page 17 presents the six-step SAGE cycle.
Official Educator Guide · page 12
The GenAI Guidance Levels table includes a SAGE Student Orchestration Level column and identifies the table as adapted from Elkhodr and Gide.
Official Educator Guide · page 17
The Educator Guide presents Generate, Evaluate, Refine, AI Critic, Reflect and Defend as the six steps of SAGE.
Structured AI-Guided Education (SAGE) is an evidence-informed pedagogical framework designed to develop students’ capacity to use generative AI critically, transparently and accountably.
SAGE does not determine whether GenAI is permitted in an assessment. Permission remains governed by the relevant institution, course, unit and assessment instructions.
Where GenAI use is permitted, SAGE provides a structured process through which students generate, evaluate, refine, critique, reflect on and defend AI-supported work.
Students begin from an informed human starting point and use GenAI to generate material, possibilities or perspectives for examination. The purpose is not to accept the first output as an answer, but to create material that can be interrogated through the subsequent stages.
Students evaluate AI-generated material against authoritative disciplinary evidence. They identify inaccuracies, unsupported claims, omissions, bias, inappropriate recommendations and areas requiring further investigation.
Students decide what to accept, modify or reject. Refinement is based on disciplinary evidence and human judgement rather than simply asking the AI to improve its own output.
Students deliberately use GenAI as a critic or challenger of the developing work. The student then evaluates the validity of that criticism rather than assuming that the AI-generated critique is itself correct.
Students consider how GenAI influenced the work and their learning. Reflection should identify where AI was useful, where it was unreliable or limited, what human decisions were required, and what the student learned through the process.
Students demonstrate that they understand, can justify and can apply the reasoning represented in their work. Defend provides an opportunity for personally attributable demonstration of understanding rather than relying solely on the submitted artefact.
The SAGE Student Orchestration Levels describe the observable quality of a student’s collaboration with generative AI. They are developmental capability levels, distinct from guidance or permission levels.
Accepts AI output with minimal critical evaluation, disciplinary verification or justification.
Filters AI output using disciplinary knowledge and evidence, accepting, modifying or rejecting selected elements.
Systematically combines human and AI contributions through explicit trade-off reasoning, contextual judgement and iterative refinement.
Proactively identifies gaps, assumptions and bias, then develops a traceable synthesis or contextual adaptation that extends beyond both the original human position and the AI output.
SAGE does not treat GenAI output as evidence.
AI-assisted work should be evaluated against an authoritative anchor appropriate to the discipline and task.
The anchor provides students with an external basis for testing AI-generated claims rather than asking the AI to verify itself.
This supports one of the central SAGE principles:
An anchor may include:
SAGE treats assurance as something that can be distributed across selected points in the learning process and mapped to the learning outcomes that require confirmation.
Rather than assuming that every activity requires the same level or form of verification, distributed assurance asks where meaningful evidence of a particular capability can be gathered as learning develops.
The columns represent selected points in a learning sequence, not a weekly assessment schedule.
| Learning outcome | Early learning | Developing | Later learning | Confirmation |
|---|---|---|---|---|
| Outcome A | Corroborative | Controlled | Direct | |
| Outcome B | Corroborative | Direct | ||
| Outcome C | Controlled | Direct |
For example, a presentation may provide process evidence, operate under controlled conditions, or become a direct assurance task through individual learning-outcome-mapped questioning.
SAGE AI records, source trails, accept–modify–reject decisions, portfolio or version histories and reflections make judgement more visible. They strengthen the evidentiary picture but do not alone confirm independent capability.
Supervision, restricted computing environments, identity verification, controlled windows, personalised variants or supervised partial completion strengthen attribution. The condition is not the capability; the evidence produced remains what matters.
Interactive oral assessment, outcome-mapped questioning, practical demonstration, supervised challenge, scenario defence or live walkthrough and debugging can directly elicit capability. Defend is the SAGE design principle for this function, not a prescribed format.
Selected assurance points can be positioned across an assessment sequence, unit or course. A learning outcome may accumulate corroborative evidence at one point, encounter controlled conditions at another, and be directly demonstrated where stronger confirmation is warranted.
A blank assurance point simply means that the learning outcome is not being targeted for assurance at that point. It does not imply that the learning activity has no pedagogical value.
When assurance is concentrated in a small number of high-weight events, evidence of capability is correspondingly concentrated. Distributed assurance provides an alternative architecture in which appropriate evidence is gathered at selected points as learning develops.
The objective is not to maximise surveillance or assessment events. It is to build sufficient, proportionate and learning-outcome-aligned evidence of capability across the learning process.
A detailed overview of the framework, its six-step cycle, pedagogical rationale, implementation principles and supporting evidence.
Explore the frameworkA student-facing guide to beginning from personal understanding, evaluating AI output, using authoritative anchors, documenting decisions, reflecting on AI use and preparing to Defend the work.
Open the Student GuideAn educator-facing interactive resource supporting the design of a proportionate Defend checkpoint based on the capability, cohort and teaching context.
The tool supports pedagogical planning. It does not determine or override institutional assessment requirements.
Open the Defend ToolAn eleven-lesson interactive module introducing higher education students to responsible generative AI use, critical evaluation and AI-supported learning strategies.
Two lessons, Cognitive Learning Strategies and GenAI Mirroring, were contributed by CQUniversity’s Academic Learning Centre.
Explore GenAI-101A permission-to-pedagogy companion that sets clear expectations for responsible AI-supported authorship, verification, disclosure, reflection and defensible understanding.
Explore SAGE-APThis section introduces related SAGE resources that educators may find useful.
Editable guides, planning resources, student materials and other SAGE resources available for educators to adapt within the applicable licence and institutional context.
Browse downloadsResearch, publications, sector engagement and evidence associated with the development and evaluation of SAGE and related approaches to responsible GenAI-supported learning.
View evidence and publicationsA developing community for educators, researchers and learning designers interested in responsible, evidence-informed approaches to GenAI-supported learning and assessment.
Explore the communitySAGE continues to develop through research, classroom implementation and dialogue with educators, students and learning designers.
If you have an idea, suggestion, implementation experience or reflection that may contribute to the development of these resources, we would be pleased to hear from you.
Educators and researchers interested in ongoing discussion and collaboration are also invited to consider joining the SAGE Community of Practice.
CQUniversity. (2026). Guided–assured assessment model educator guide. Learning and Teaching Futures, CQUniversity Australia.
This independent page is maintained by the authors of the SAGE framework. It brings together the places where the CQUniversity Educator Guide references SAGE and connects readers with the wider SAGE framework and educator resources.
The CQUniversity Educator Guide is the source for the Guided–Assured Assessment Model; the SAGE website provides the independent framework, guidance and resources.