Teaching and Generative AI
Navigating Generative AI: Six Suggestions for Every Instructor
The following suggestions have been prepared by the Centre for Teaching Support & Innovation based on engagement with U of T instructors and current recommended practice. As you consider the impacts of generative AI on your teaching, you may wish to respond by:
- Clarifying expectations with your students by discussing your expectations and providing guidelines around using generative AI tools in your course. Add clear language to your syllabus and assignments regarding allowable use.
- Preparing for a conversation with your students about responsible use of generative AI for learning in relation to your course and discipline.
- Rethinking both learning outcomes and corresponding assessments with the potential impacts of use by students in mind. Take time for critical consideration of teaching with generative AI.
- Talking to your TAs about expectations for use of generative AI in relation to their role and to your expectations for appropriate use/non-use by students in the course. Consider sharing TATP’s TA-focused resource on generative AI with your TAs.
- Familiarizing yourself with tools that align with the University’s privacy and data protections. If leveraging the capability of generative AI, you can use Microsoft Copilot in Protected Mode to protect your data and privacy.
- Exploring applications of generative AI tools and their outputs to gain a better understanding of their capabilities and limitations. There are a number of workshops and resources available through the Centre for Teaching Support & Innovation.
For more information: https://ai.utoronto.ca/faculty/
Download this resource: PDF | Word | PPT
For more downloadable GenAI resources: CTSI Resources
Share with Your Students: A Companion Resource
Using AI Tools for Learning at U of T is a student guide from the Centre for Learning Strategy Support (CLSS), offering six key tips for responsible, effective, and ethical GenAI use.
Planning Your Assessments
1. Developing or Revising Learning Outcomes
Learning outcomes describe the knowledge or skills students should acquire by the end of a class, course, or program. They are not fixed: they may shift to reflect your discipline, the requirements of follow-up courses, students’ career paths, and the digital literacy skills students will need as AI becomes more common in future work.
As you decide whether and how to engage with generative AI, you might consider: which skills you want students to develop and how to express them as outcomes; whether AI use can align with your outcomes and teaching philosophy; how AI might be used to deepen thinking; and which digital literacy skills matter for your students.
1.1 Foregrounding Human-Centred Skills
Some cross-disciplinary cognitive skills are worth foregrounding as AI tools become more capable. Oregon State University’s “Bloom’s Taxonomy Revisited” (Figure 1) reconsiders what meaningful learning looks like given AI’s capabilities. At the “analysis” level, for instance, generative AI is already proficient at comparing, contrasting, and inferring themes—so you may want to reframe outcomes at this level toward skills that do not invite overreliance.
To check whether an outcome targets human-centred skills, consider whether it asks students to interpret authentic problems and choices, to engage higher-order thinking (critical analysis, synthesis, evaluation), and to develop conceptual knowledge—the “why” behind the “what.”
Bloom’s Taxonomy Revisited: A framework for aligning course activities and assessments with higher-order thinking skills, updated to reflect the evolving role of generative AI in education. Oregon State University Ecampus, CC BY-NC 4.0
1.2 Including Learning Outcomes That Encourage AI literacy
AI literacy does not require students to use or prompt AI tools; it focuses on being a critical, responsible user of AI-generated content. The University of Toronto Libraries’ Framework for AI Literacy offers a discipline-adaptable structure built around three frames that learners move through fluidly:
- Understand: how AI tools work, how they are trained, and the ethical, legal, and social implications of their use—including privacy, bias, labour, and copyright.
- Use: engaging with AI tools responsibly and effectively, in line with relevant policies and an awareness of data and privacy considerations.
- Evaluate: judging credibility and accuracy, and the broader impacts these tools have on learning and on the world.
Your liaison librarian can help tailor AI literacy instruction to your discipline, and U of T’s GenAI Literacy Course Modules offer customizable, ready-to-use content.
2. Designing Generative AI Out — to Keep Student Thinking Visible
Once you have revisited your course-level learning outcomes, the next step is designing or revising your assessments to match. A useful shift here is moving from “Can students use generative AI?” to “What learning needs to remain visible?” The Digital Education Council (2025) calls this AI-resilience: rather than relying on a “No AI” rule and hoping students comply, you structurally design tasks so the skills at the core of your outcomes can’t easily be outsourced — which takes redesign, not just rules or detection tools (Corbin et al., 2025).
There are a few ways to build that resilience: you can design generative AI out of an assessment where independent thinking must stay visible, design it in where it can support learning, or validate it across a sequence of assessments rather than within just one.
For assessments designed to develop or test students’ unaided thinking and foundational skills, the most reliable approach is to make AI use structurally difficult or unnecessary, rather than prohibiting it and hoping for compliance (Digital Education Council, 2025). Consider the following strategies:
- Human-centred skills: Focus assessments on evaluating skills that generative AI cannot easily replicate, such as critical analysis and creative problem-solving.
- Traceable process: Require intermediate artifacts—outlines, drafts, revision notes, or annotated planning documents—submitted alongside the final work, so the evolution of a student’s thinking is visible and individual to them. Metacognitive components (“Describe the reasoning behind your approach”) make that thinking explicit. See CTSI’s Assessment Process Reflection Template.
- Synchronous and supervised formats: Where appropriate for high-stakes or summative moments, shift from asynchronous to synchronous tasks—in-class writing, oral exams, live presentations, or classroom discussion—which are structurally resistant to AI interference because they remove access during task performance.
- Local and personal context: Focus prompts on material specific to your course—recent in-class discussions, current events, course-specific data—or ask students to link course concepts to their own experiences, so responses cannot be generated without the context students actually have.
U of T Instructor Example: Reading Annotation Assignment (No GenAI)
S. Trimble, Assistant Professor, Teaching Stream & Associate Undergraduate, Women and Gender Studies Institute, UTSG
As part of the third-year course, Playing, Sports, Cultures (WGS331, Winter 2025) this assignment requires students to engage deeply with one assigned reading. Using the Hypothesis annotation tool on Quercus, students annotate the text with comments and questions that address both the content (“what” the article says) and the form (“how” it says it). The assignment is designed to prevent generative AI misuse by emphasizing a process-driven approach: students must demonstrate individualized, critical engagement by identifying main arguments, defining key concepts, raising questions, and making personal or course-specific connections. This focus on unique analysis and transparent process limits the usefulness of AI-generated content and supports academic integrity.
For more details on this assignment, visit Professor Trimble’s assessment example page.
3. Designing Generative AI In — to Build AI Literacy and Judgment
When generative AI is permitted, the same principle applies: determine which parts of the task require students’ independent work, and design those elements to remain resilient to inappropriate AI use, while making the human contribution — and the reasoning behind it — the focus of assessment (Digital Education Council, 2025).
- Align with learning outcomes and skill development: Design assessments where generative AI use supports course-specific learning goals and AI literacy. Consider allowing students to use AI tools for studying, brainstorming, editing, and code debugging as part of building these skills.
- Real-world and complex applications: Integrate generative AI in ways that mirror professional practice and challenge students with multifaceted problems requiring human judgment and creativity.
- Develop critical analysis skills: Ask students to critique and improve AI-generated content — including AI-generated examples or explanations — and clearly communicate how their ability to evaluate and use AI outputs will be assessed.
- Use AI-based simulations: Create low-stakes scenarios where AI plays a role — a mentor, a character, or a stakeholder — for students to practice applying course concepts. This gives students a space to test their thinking and get feedback before a higher-stakes assessment.
- Encourage reflection: Ask students to submit written reflections on how they used generative AI, analyzing its impact on their learning and approach. See CTSI’s Assessment Process Reflection Template.
- Citation requirements: Share citation resources (such as University of Toronto Libraries “Citing Artificial Intelligence (AI) Generative Tools (including ChatGPT)” guide) and incorporate citation requirements for generative AI use, including the prompts used.
U of T Instructor Example: AI-Integrated Concept Map Activity
Nirusha Thavarajah, Associate Professor, Teaching Stream, Department of Physical & Environmental Sciences, University of Toronto Scarborough
As part of a CHMA10 (Introductory Chemistry) lab on determining acetic acid content in vinegar, Professor Thavarajah had students use Microsoft Copilot to visualize titration procedures through concept maps, iteratively refining and documenting their prompts. A required 250–300 word reflection — completed without AI — asked students to evaluate Copilot’s accuracy and limitations, and what the process revealed about their own learning. The assignment builds disciplinary understanding alongside AI literacy: prompt engineering, critical evaluation of AI output, and awareness of when AI genuinely supports learning.
For more details on this activity, visit Professor Thavarajah’s teaching example page.
4. Validating Across Assessments — Not Just Within Them
Not every assessment can be made fully AI-resilient on its own, and it does not have to be. Where a single task is difficult to protect, validity can instead be established across a connected chain of assessments, with each task building on the student’s earlier work in a way that is contextual to them. In this model, confidence in student learning comes from the coherence and progression across tasks rather than from any single submission (Digital Education Council, 2025). This shifts the picture of learning from a one-time snapshot toward a pattern of growth over the term, and pairs well with distributing evidence across several smaller, varied touchpoints rather than concentrating it in one or two high-stakes products.
It can also help to consider that generative AI in assessment may be a “wicked problem”: one that resists a single correct solution and is context dependent (Corbin et al., 2025b). Framed this way, the goal may not be a perfectly AI-proof assessment, but a workable balance between what is meaningful to assess and what is sustainable for you and your students. There is no need to immediately make every assessment fully resilient, or to redesign everything at once. The difficulty sits in the problem itself: adapting over time, and weighing teaching goals against workload, is sound practice.
5. Designing Inclusive Assessments with GenAI in Mind
Universal Design for Learning (UDL) is a proactive design approach that reduces learning barriers and welcomes learner variability. Applying UDL principles — whether or not you use generative AI in assessments — helps you anticipate barriers rather than react to them, and builds student agency in the process. UDL doesn’t replace specific accommodations, but it helps close gaps between diverse student needs and instructional design.
Three questions can guide this work at any point in your assessment:
- What learning am I trying to make visible? Name the specific skill or thinking the assessment should reveal, not just the product students hand in. This determines which parts of the task students must do independently — and where generative AI is designed in or out.
- What barriers might interfere with students showing that learning? Barriers are features of the task, not student deficits: unclear instructions, an unfamiliar format, one narrow mode of expression. Ask whether the assessment measures the learning you named, or something incidental to it. Generative AI can lower some barriers (clarifying instructions, helping plan a first step) and raise others (producing the thinking you meant to assess).
- What learner variability am I planning for? No single mode of expression works for all learners. Students also arrive with uneven access to and confidence with AI tools. Where multiple modes could show the same learning, offer options (written, visual, oral, multimodal). And make scaffolding — exemplars, annotated rubrics, practice materials — available to everyone, rather than leaving students to find that support privately through AI.
Communicating Expectations
1. Setting Clear Expectations for Student GenAI Use
After finalizing your learning outcomes and assessments, consider how you will establish and communicate clear course policies on generative AI. Students bring different levels of familiarity with AI tools and will look to you for guidance on permitted uses and their impact on learning.
The first day of class is a key opportunity to set this tone — modeling how you hope and expect the course will proceed. Building a sense of community through active participation around your policy, rather than simply presenting it as a fixed rule, helps establish expectations that support responsible generative AI use throughout the term.
The University of Toronto has created sample syllabus statements you can adapt for your syllabus and assignments to clarify what AI use is or isn’t allowed. Beyond the syllabus itself, plan how you will reinforce these expectations in class and on Quercus. Consider one or more of the following strategies:
- Explain your policy and its rationale: rather than simply repeating your syllabus language, explain why you chose this policy and how it supports the course’s learning outcomes.
- Create a community agreement: ask students to collaboratively reach consensus on generative AI in course activities and assessments. See the CTSI resource for more guidance on building these agreements.
- Create space for discussion and reflection: before addressing your policy directly, open a broader discussion on learning and generative AI. Active learning activities — low-stakes writing, think-pair-share, jigsaws — are effective ways to generate and record student thinking.
- Outline best practices for using and citing generative AI, if it is permitted in your assessments.
- Point students to learning strategy and writing support: at UTSG through CLSS and the Writing Centres; at UTM via the Student Resource Hub; at UTSC through the Centre for Teaching and Learning and Academic Advising and Career Centre. These offer appointments, workshops, and peer coaching.
U of T Instructor Example: Co-Constructing a GenAI Course Policy
Daniel Corral, Assistant Professor, Department of Leadership, Higher and Adult Education, Ontario Institute for Studies in Education (OISE)
LHAE 5815 (Postsecondary Finance and Accountability) is an online, synchronous graduate course focused on how postsecondary institutions serve historically underrepresented students. Rather than arriving to class with a pre-written GenAI policy, Professor Corral developed the course policy through a structured co-construction process with students grounded in metacognitive principles and students-as-partners pedagogy. The process moved from sharing institutional guidance, to small-group discussion of fair and ethical GenAI use, to a refined final policy distinguishing between GenAI as a learning aid (clarifying concepts, editorial support) versus uses that bypass learning (generating ideas or written content). The resulting policy includes assignment-specific guidelines in an accessible table format and a reflection requirement for any permitted GenAI use.
For more details on this approach, visit Professor Corral’s teaching example page.
2. Modeling Responsible GenAI Tool Use
If you are encouraging or allowing generative AI use, consider opening the course with a demonstration of relevant, institutionally approved tools. These tools can be engaging for students, but using them well takes time and practice for both instructors and students.
Since students’ familiarity with generative AI varies widely, consider one or more of the following:
- Demonstrate prompt writing directly: rather than a general AI intro, model how students could responsibly use these tools for upcoming assessments — spend focused time on what makes an effective prompt (natural-language text describing the task you want the AI to perform). See our Tool Guide’s “How can I prompt with Copilot?” for more.
- Create space for experimentation and peer feedback: the first class is a good opportunity to gauge students’ familiarity with these tools. Beyond your own demonstration, encourage students to share tips and reflections as they experiment independently.
- Point students toward academic supports: model how to connect with writing centres, learning strategists, and other supports across U of T — even if they’re already listed in your syllabus. Students who know what’s available are more likely to use permitted AI tools responsibly.
Resource: GenAI Literacy Course Modules
The GenAI Literacy Course Modules are flexible, open educational resources to help instructors introduce GenAI literacy into their courses — use them as-is, adapt them to your context, or import them directly into Quercus. Available as importable Quercus modules, a standalone open-access shell, or downloadable PowerPoint/Word/PDF files, with an instructor guide for integration.
3. Clarifying GenAI Expectations for Teaching Assistants
When connecting with your course TAs during your initial meeting, it is good practice to clarify whether and how your course will engage or limit generative AI use. Consider discussing one or more of the following:
- Grading and rubrics: whether and how rubrics will be adjusted to account for generative AI, so human skills stay the focus of evaluation. Ask TAs about their own grading experience since AI tools became widely available — their insights can inform your protocol.
- Student communication: how TAs should communicate your course’s GenAI policies and recommended practices to students.
- Tool training: if generative AI tools are part of the course, train TAs on their use — including how to model responsible use if their role involves tutorials, labs, or office hours.
- Check-ins: a plan for ongoing communication, so TAs have a clear way to raise questions or concerns as they come up.
See TATP’s Teaching with GenAI: Considerations for Teaching Assistants for further guidance you can share with your TAs.
Example: Instructor-TA Team Meeting Plan
The initial instructor-TA meeting typically involves a review of the Description of Duties and Allocation of Hours (DDAH) forms. This is also a good opportunity to clarify responsibilities and communication protocol around generative AI. Consider organizing the conversation around:
- How should TAs handle student questions about generative AI use in the course?
- What’s the protocol if a TA suspects unauthorized AI use in an assessment?
- How can the instructor and TAs work together on any course-size or capacity challenges generative AI raises?
4. Explaining Assessment Rationale and Procedures
Since generative AI use varies across courses, students need clear guidelines — not just stated once, but reiterated throughout the term. A few ways to do that:
- Communicate the value of assessments: connect what students gain from an assignment to the course’s learning outcomes, including how (and why) generative AI fits in if it’s part of the task. TAs running tutorials can help reinforce this.
- Review assignment instructions in class: clearly explain how much AI use is allowed, and talk through why you chose to encourage or discourage it — and where it may or may not actually help with the assignment.
- Include an integrity statement or reflection form: giving students space to reflect on the resources they used can support their adherence to your GenAI policy.
- Share a “ready to submit” checklist, covering four things: the purpose (skills or knowledge gained), the task (what to submit, how, and by when), the criteria (what’s expected), and the process (when and how AI may support the work). A clear checklist like this helps students understand the value of the effort involved — which makes them more likely to use AI responsibly rather than as a shortcut (Bowen and Watson, 2024).
Resource: Assessment Process Reflection Template
CTSI’s Assessment Process Reflection Template helps students document how they completed an assignment, including their use of AI and other supports. You can use it as a graded component, a completion-based activity, or a standalone reflection, and adapt it for any course or platform.
Students answer three questions: what tools, resources, or people they used and how they credited them; why they chose each resource and how they used it; and their process for planning, drafting, and revising.
The fourth part is a set of supporting materials, like drafts, notes, or AI prompt logs, submitted alongside the reflection. This works well as a small portfolio of the student’s work, with the reflection written as a kind of cover letter to it. It can help students take ownership of their work and make their thinking visible, not just the final product.
Proactive Strategies to Address Unauthorized GenAI Use
Instructors can promote academic integrity through the ethical and effective design, delivery, and assessment strategies (Eaton, 2024). In the context of generative AI (GenAI), consider:
1. Fostering Honesty via Course Design
- Clear syllabus statements: Explicitly define authorized/unauthorized GenAI use for your course. If you need language about GenAI for your syllabi, U of T has examples available.
- Purposeful assignment design: Create meaningful assessments that encourage students to focus on the process of learning, prioritizing critical thinking, reflection, and classroom-specific content; see CTSI’s Teaching and Generative AI.
- Scaffolded assignments: Break projects into smaller components with checkpoints to identify and assist struggling students early. This encourages students to focus on the learning process, rather than the final outcome; see CTSI’s Teaching Resources.
- Reflections on AI use: Ask students to explain if and how they integrated AI tools into their workflow. If none were used, have students describe their research, analysis, and creative process. This metacognitive practice can build intrinsic motivation; see CTSI’s Assessment Reflection Template.
- For practical examples of assessment designs from U of T Instructors that use scaffolding, reflection, and course-specific application to emphasize and evaluate students’ original thinking, see U of T Teaching Examples.
2. Educating Students on Best Practices
- Clarify expectations: Collaborate with your TAs to provide ongoing discussions of appropriate vs. inappropriate uses of AI tools; see TATP’s Teaching with GenAI.
- Model proper citation: Demonstrate how to properly cite AI when permitted on an assessment; see UTL resources on citing AI, image research, and copyright considerations.
- Emphasize the value of original thought: Encourage students to recognize that their unique voice, creativity, and critical thinking are invaluable and irreplaceable when completing assessments, whether they be writing, coding, or multimedia projects.
- Discuss AI limitations and risks: Explain the limitations of GenAI (hallucinations, fabricated references), emphasizing the importance of fact-checking and digital literacy.
3. Identifying Possible Misconduct Cases
- Before issues arise, familiarize yourself with traditional detection methods (e.g. the student cannot explain their work) and the standard academic misconduct process.
- AI-detection software programs are unreliable and biased against non-native English writers (Elkhatat et al., 2023; Liang et al., 2023; Saha and Feizi, 2025). U of T does not support the use of AI-detection tools; see the Office of the Vice-Provost, Teaching & Learning (OVPTL)’s FAQ on Generative AI.
- Personal intuition that a text is AI-generated has been shown to be inconsistent, even when evaluators are experienced with GenAI and confident in their abilities (Waltzer et al., 2024).
- For further guidance, contact the head of your academic unit.
Download this resource: PDF | Word | PPT
For more downloadable GenAI resources: CTSI Resources
This work is licensed under a Creative Commons BY-NC-SA 4.0 International License.