There’s a lot of data out there now on student uses of AI, but the best data is your data. This post summarizes general data on student uses of AI, then provides survey questions, study methodologies, and ideas for learning more about your students and AI. Knowing how your students are engaging with AI is part of designing AI-aware courses and assignments.

A raft of recent surveys, articles, and studies have taught us a lot about how undergraduate students are using generative AI, what they feel about that use, and what they think of AI in general. Surveying these studies over a year ago, I noted a few trends: most students were using generative AI in some capacity; use varied across discipline, gender, and socioeconomic status; and students had mixed feelings about their future with AI along with the preparation their institutions provided for this future. My own research indicated that students also felt AI was fraying their human relationships at school: they sometimes sought help from AI rather than their instructors, and they resented their peers who overused AI, but the surveillance measures faculty were taking to mitigate that use were undermining their mutual trust with instructors. A small, but significant number of students admit using AI to cheat.

In the past year, we’ve seen students increasingly resistant to AI on environmental and political grounds, and they are more concerned about their own potential cognitive offloading to AI. Yet most of them still use it regularly to help understand complex material, save time, and improve their grades. Reviewing the studies and my own research, I estimate that only 5-15% of students refuse to use AI altogether. Those who don’t refuse often feel guilty, resigned, or forced to use AI to compete with peers.

Students’ AI environments are also rapidly changing: more institutions have signed enterprise agreements that provide them free access to frontier AI models like ChatGPT and Claude, Copilot or Gemini. Some of these companies have offered educational access to students directly, often advertising their models on campus. Google’s AI Overview now appears in the majority of searches on their site, and other platforms are relentlessly integrating AI, which means it is more difficult than ever to avoid generative AI altogether.

This research paints a complex, if relatively consistent picture of student AI use overall. For instructors, it’s helpful to know these general trends of student AI use in order to design AI-aware courses and curriculum. But each of these studies have a delay from data to reporting (sometimes two years!) and the AI landscape for students is changing rapidly. Additionally, available data overrepresents American undergrads at research universities. If you teach at a community college, high school, faith-based institution, HBCU, Hispanic-serving institution, SLAC, or a school outside the US, your students’ voices, behaviors, and attitudes towards AI are underrepresented in this research.

Besides, every school is unique! And high variability across disciplines means that public data may not speak to the specific experience of students in your department or program. The best data is recent, local data. The rest of this post gives you tools to collect that data on your own students so that you can calibrate your own, local AI-aware teaching strategy.

What do you want to know about your students?

When ChatGPT was released in November 2022, I was directing the University of Pittsburgh’s Composition Program, which serves about 7,000 students annually through first-year, gen-ed, and specialized writing courses. Since my research was about technologies of writing, and I’d previously argued we should be paying attention to developments in AI, I was interested in how students were interacting with this new tool that could write. I quickly wrote up a survey and contacted composition instructors to send it to their students at the end of the Fall 2022 semester. I wanted to know if students were aware of ChatGPT, if they were using it, and how they thought such a tool should be used in writing classes.

My Fall 2022 survey was answered by 77 students. Their answers taught me a few things that were helpful at the time: most students didn’t know about ChatGPT yet, but they didn’t think that computer programs should be used to write full paragraphs or essays in their composition classes. About half of them thought it was okay for such a program to “fix errors, formalize the tone, and clean up the style.” No students reported using ChatGPT in December 2022 and answers to the open-ended questions indicate that their primary reference for such tools was Grammarly. But by Spring 2023, 25% of students had tried using ChatGPT, and 40% had heard of it. By Fall 2023—a year after ChatGPT’s release—almost all students (97%) had heard of the technology and most (62%) had tried using it. Students’ open-ended responses suggested that they had developed opinions about these tools as well. AI was, variously: helpful, plagiarism, and crazy.

I’ve given this survey in various forms almost every semester since then, adding questions about how AI tools affected their motivation to write, whether they thought their instructors should use AI, and why they choose to use AI or choose not to use AI. I have shared key results with composition instructors via our listserv and teaching cluster meetings. These results also informed our development of suggested policies for the program. We also used results to design questions for a focus group study in Spring 2025, which I’ll say more about below.

Figures 1 and 2 below show the initial breakdown of students’ thoughts on how computer programs should be allowed to help with writing in composition classes and how remarkably stable that distribution looks over the Fall 2022 – Fall 2024 semesters. Figure 3 shows the results of questions I asked about motivation to write (questions I borrowed from colleagues at Carnegie Mellon University). Over the year and half from Spring 2023 to Fall 2024, AI tools appear to have decreased students’ motivation to write significantly.

FIGURE 1: Students don’t think it’s okay for a computer program to write full essays

FIGURE 2: Students grew more comfortable over time with computer programs helping to revise essays

FIGURE 3: How students feel about AI writing tools, Spring 2023 to Fall 2024

Students’ access to AI tools, their fluency with them, attitude towards them, feelings about their peers’ uses of AI, and experiences across different classes at an institution all shape the landscape of AI for teaching. You can make more informed choices about policies, assignments, and curriculum if you know about your students’ approaches to and ideas about AI.

What would you (and any collaborators) most like to know about? Consider the following questions and choose a few to focus on.

• How are students using generative AI in their classes?

• How comfortable are students with using generative AI?

• Where do students think it’s okay to use generative AI, and where would they refuse to use it?

• Have students used generative AI when their instructor or generative AI policy disallowed it?

• What concerns students most about generative AI? What excites them most?

• What kind of support do students want for generative AI from the institution?

• How is the presence of generative AI affecting students’ academic relationships with each other and their instructors?

• How do students feel about the impact of generative AI on their learning and cognition?

• What do students think about the institution’s generative AI strategy?

• How do students imagine their future with generative AI?

Remember that due to survey fatigue, the longer your survey, the fewer students will complete it. Choose questions that reflect what feels most relevant to your pedagogy or program at the moment. You can always follow up with different or revised questions next semester.

How will you learn about your students and generative AI?

After you’ve decided what you most want to know right now, you’ll want to figure out what method to use to learn about your students and generative AI. Questions that are better answered anonymously are a good fit for surveys. Questions that get at feelings about people or futures work better in conversation such as interviews or focus groups. There are many ways to connect with your students. How you do so will depend on what you want to learn, any collaborators or sponsors you recruit, and whether you have funding for your work.

Do you have access to funding? If so, then you could use it to pay students to participate in interviews or focus groups, to lead discussions, or to host an event with food. Funding might also be helpful for data analysis, visualization, or communication, if those aren’t skills you have. Consider employing undergraduate interns or researchers if your institution has a program to support that. If your department chair or teaching center can offer you a release of teaching or service or a stipend for carrying out your research, then take it! It can’t hurt to ask at any rate.

What support can your campus partners provide? Before you start your research, check in with your teaching center, provost office, advising, or IT department. They may have statistics on students’ AI uses, examples of AI policies across campus, or information about academic integrity violations. They may also be able to provide you with support on your study design and distribution. A few semesters into my surveys, I reached out to my own Center for Teaching and Learning and got great advice on how to consolidate and clarify questions. My student focus groups were sponsored by our IT department, Pitt Digital. Other teachers and administrators may also be able to help with research design, distribution, and funding. And your institution’s communications department may be able to help translate or publish information about your study to a wider audience or to your peers and students. For example, my university’s communications team wrote up an overview of how students are navigating AI, and featured a video the Pitt Digital comms team made.

What kind of approval do you need to conduct your research? Another reason to check in with your department, teaching center, and administrators is to learn whether you’ll need permission to conduct your study. Is it okay to survey students through their composition classes, or do you need to reach them outside of class? Can you recruit them through their courses, or do you need to put up flyers with QR codes? Ask around to make sure your research is welcomed. Additionally, most institutions have an Institutional Review Board (IRB) that governs any research conducted by its affiliates; you should check with your IRB to see if your research requires their oversight. Each IRB is different, but they typically evaluate studies for ethical considerations, privacy, data security, and research design. There are general exceptions for their full review, and they might consider your study “exempt” from IRB review if its primary goal is program improvement.

Below are a few methods that researchers have used to learn more about their students and generative AI.

• Surveys: This is the easiest, fastest, and cheapest method for learning about your students and generative AI. You are welcome to adapt my own survey questions, from Fall 2024 or Spring 2025. Surveys can include a mixture of question types including: yes/no, choose all that apply, multiple choice, Likert scale, and open-ended. You can distribute an anonymous survey through instructors for classes, departments, or QR codes on campus flyers. Students generally don’t expect payment for answering a survey, especially if they’re given class time to respond. If you have the right approvals, you can collect good data within a week. For quick analysis of results, in my experience, the more sophisticated generative AI models are pretty good. Especially for open-ended questions, AI can be helpful to get a sense of the general pulse of responses. AI analysis won’t be accepted for research publications, however, and your IRB may not allow it. At the very least, be sure that any identifying information is stripped from your data prior to AI analysis and use models that your institution provides, if you have access to those.

• Student interviews or focus groups: You can get students together in one-on-one interviews or in focus groups of 3-5 to answer a serious of questions. This method can provide more nuance to what you learn from students, especially if you ask follow-up questions. Students can learn from each other in focus groups, which also helps the conversation be less awkward and formal. It’s hard to keep students anonymous in this research method, but you can avoid collecting identifying information to protect student privacy. Plan on conversations under an hour and record at least the audio of the discussion (avoiding video also helps protect privacy). You can transcribe the conversations (or have AI do it through Zoom, Otter.AI or Whisper) and analyze them in more depth. You can adapt the interview script the University of Pittsburgh team used in Spring 2025 or come up with your own questions to guide the conversation.

• Student-faculty workshops: Gather students and faculty together for a conversation about AI by asking faculty to bring their classes, advertising through the library, or inviting them to a workshop where you provide food or extracurricular credit. When I run workshops like this, I’ve often broken the ice by having students and faculty work together to answer questions on a popular AI benchmarking test. In your own context, feel free to use my quiz, You vs. MMLU, which was based on my colleague Matt Burton’s original design. In the workshop, you can break students and faculty up into small groups to discuss and share out. Or you can collect real-time responses from students and faculty using a QR code and a service such as Mentimeter. Using Mentimeter, I like to ask both students and faculty to answer questions about AI use and what they’d like to tell each other. Mentimenter allows an audience to use their phones to answer multiple choice and open-ended questions, and you can display the results in real-time on a screen.

• Student-led conversations: In Spring 2026, Boston College developed an innovative way for students to facilitate conversations amongst each other. In the “BC Students Talk AI” program, students signed up for a slot, ate and talked together, then submitted reflective feedback to the faculty sponsors. Organizers provided a different set of materials to faculty, which could support “plug and play” conversations within courses. The organizers write,

“Our goal is to create space for deliberation, discernment, and formation. We see this space as one in which students will reflect on how their values, disciplines, and responsibilities shape their use of emerging technologies.”

The BC conversations had strong uptake among students: 47 students hosted conversations with 242 students total; 7 instructors integrated the conversations into their classes; and the organizers received 96 responses to analyze from these conversations. In these responses, it became clear to the BC team that students have nuanced, ambivalent views of AI, just like faculty, and they are eager for more spaces to talk about it with each other and with their professors. Visit the website for more details on the “BC Students Talk AI” project.

Continuing the conversation

Conversations with students can help faculty design training and curriculum that better meets students’ needs but also break down traditional silos as they collaborate to design this support. For example, in a recent Substack post analyzing three large-scale studies of what students say about generative AI, Tawnya Means argues that the support institutions are providing students for generative AI doesn’t align with what they’re requesting:

“Students are not asking to be trained as AI users. They are asking to be developed as thinkers who happen to have new instruments available. These are two fundamentally different educational projects, and most institutions have defaulted to the first because it is easier to operationalize. The cost of that default is significant — and it shows up in the data.”

Better, local data can help us better align our programs to our students’ specific needs. Check out the Conclusion to the The Norton Guide to AI-Aware Teaching for more on how institutions are running these conversations and collecting local data.

Wherever you are at with connecting with your students, continuing the conversation is great! You can start small by asking questions for discussion in your own class or offering an anonymous survey during class time. Take one more step by asking your colleagues to distribute the same survey and discuss results over lunch or share them with your department. You could then collaborate on drafting follow-up questions from the surveys and ask them in small focus groups or interviews with students. Or reach out to other departments to scale up your study. Small, pedagogically focused studies are great for graduate students learning to teach or conduct research, so include them in the conversation if you can.

Research is a form of teaching, too! One thing students consistently say in my own surveys, workshops, and focus groups is that they want to talk with their instructors about generative AI, and they are glad to have their ideas valued. By making the effort to ask students about their experiences and thoughts, you can not only align policies and tailor support that helps your students but you can also show them you’re listening.

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more.

Subscribe now

Thanks to Erica Luckert, Director of Composition & Assistant Professor of English at University of Southern Mississippi, for her helpful feedback on a draft of this post.

Resources

Examples and support for your research

• Norton Webinar, “How Are Students Using AI? A Research Toolkit for Faculty,” by Annette Vee, Nov 11, 2026: https://seagull.wwnorton.com/aiaware/resources#calendar-a0c261f3-5ef6-4a55-a2c8-cb738bd53508-event-ad0ca255-9b78-465a-9da8-04f220ebb372

• Annette Vee’s Research Toolkit for Student Uses of AI, including sample surveys and focus group questions: https://bit.ly/VeeResearchToolkitForStudentUsesOfAI

• Tawnya Means’s Spring 2026 review of reports on student uses of AI: “What Students Want Is Not What We Are Offering”

• Jennifer Trainor’s discussion of what she learned from surveying her own students about AI: “What Are Students Actually Looking for When It Comes to AI and Writing?”

• TeachLab podcast, “The Homework Machine,” a seven-episode series from 2025 featuring voices of students and teachers on AI. Focused on high school but relevant to college teaching: https://www.teachlabpodcast.com/

Recent studies

• Inside Higher Ed / Student Voice study from Spring 2026, which surveyed over 1000 students across institutions and includes a clear and accessible write-up of findings: https://www.insidehighered.com/news/student-success/academic-life/2026/06/11/why-students-arent-all-ai-and-what-they-want-colleges

• Pew Research, “How Teens Use and View AI,” with data from Fall 2025: https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/

• Rand report on teens and AI, “More Students Use AI for Homework, and More Believe It Harms Critical Thinking,” with data from the American Youth Panel, Fall 2025: https://www.rand.org/pubs/research_reports/RRA4742-1.html

• Brookings report on teens and AI, “A new direction for students in an AI world: Prosper, prepare, protect,” https://www.brookings.edu/articles/a-new-direction-for-students-in-an-ai-world-prosper-prepare-protect/

• A large-scale survey study of 95,000 students across 20 institutions: Chirikov, Igor, et al. “Generative AI Use and Misuse Call for Assessment Reform in Higher Education.” Science, vol. 392, no. 6800, May 2026, pp. 818–20. https://doi.org/10.1126/science.aec5115.

We’re excited to announce that The Norton Guide to AI-Aware Teaching e-book is officially out! My co-authors and I wrote this guide on a brutal timeline, and somehow it was a genuine pleasure — that’s thanks to working with brilliant co-authors Marc Watkins and Derek Bruff, and the great team at Norton. At the end of this post are the guide’s Table of Contents, plus links to access the book, listen to podcasts about it, or participate in events around it.

What I’m most excited about: this is a book about AI that’s focused on pedagogy. We draw on solid pedagogical foundations such as backward design, clear learning goals, and connecting with students to help teachers keep their focus on teaching even with the challenges and possibilities of AI in the undergraduate classroom. We have countless examples of how teachers are effectively navigating AI in physics, business, biology, literature, economics, and composition classes, plus frameworks and templates that readers can adapt.

Knowing that teachers are choosing to either resist or engage AI across different disciplines, classes, and assignments, we tried to balance our guidance along that spectrum to offer something for everyone. AI-aware teaching doesn’t dictate a particular AI stance: instead, it’s about listening to students, understanding what AI can do, and focusing on learning goals to help your students meet them now that AI is in the picture. We think it’s possible to do that by resisting AI or engaging AI or anything in between. I’m glad that, through this guide, we can offer a coherent and comprehensive approach to teaching and AI that I really believe in. I learned so much from Marc and Derek, the Norton team, and the many teachers we talked with while researching and writing this book. I think others can learn from it, too.

How this guide came about

I’ve been researching and teaching about AI and writing for about a decade now. And when ChatGPT came out, I was directing a large composition program. A couple of years earlier, like many of you, we’d gone through the pandemic together. Remember the “pivot”? I’d worked with an amazing team of teachers to navigate online teaching and everything that came with it—vanishing students, illness, fear of illness, black boxes on Zoom, disconnection. So, I won’t say I was prepared for ChatGPT. No one was! But I had some relevant experience.

As I began figuring out how to navigate this new challenge to our mission of teaching general education writing classes, I distributed a survey to our students, once again gathered super teachers, and considered how we might yet again “pivot” our program. But what I realized was: no one wanted to “pivot,” including me. But, even more than Covid had, AI felt directly threatening to our central goal of helping students learn. What we needed instead was a strategy that could accommodate a wide variety of good teaching approaches and keep the focus on student learning.

I thought that perhaps we didn’t need to fully “pivot.” We just needed to collectively be aware of AI and to adjust our teaching accordingly. I wrote a course policy that focused on responsibility in writing and offered versions that banned or allowed generative AI. I began promoting AI-aware teaching in workshops. I found that people appreciated the practicality of it: pay attention to students and AI; learn a bit about what AI does; and reconsider the path your students take to your learning goals because AI may have changed them.

Norton liked the approach, too, and invited me to author a guide for teachers on AI. They recruited the amazing Marc Watkins and Derek Bruff, whose work I admired. Marc writes a great Substack, Rhetorica, along with insightful columns in The Chronicle of Higher Education. Derek produces a brilliant newsletter and a lively podcast, Intentional Teaching, which highlights the great work of guest teachers, and which I like to listen to on my commute. It turns out that we like working together, and I think we made a great team working on this guide. One of the take-home points of the book is: we can do this if we work together! In the book’s concluding chapter, we offer advice on collaborating with colleagues.

If you’re interested in reading the guide or learning more about it, we have a bunch of podcasts and events for you! (with credit to my co-author Derek for compiling this info.)

To access the book:

• Norton is offering the guide as a free e-book for all instructors currently adopting a Norton textbook. If that’s you, you can contact your local Norton representative for instructions on how to access.

• If you’d like the e-book but aren’t a Norton customer, you can order it through Amazon and Barnes & Noble.

• To pre-order the paperback version of the book, visit Norton, Amazon, and Barnes & Noble and receive when it’s released Sept. 24. If you order through Norton, use code AIFREESHIP at check out to get free shipping.

• If you would like to order multiple copies for a campus reading group or some other faculty development effort, Norton has an option for you: On orders of 10 or more print copies, we offer 50% off the list price and free domestic shipping. (Such orders must be on a nonreturnable basis.) To take advantage of this offer, contact Peter Wentz at pwentz@wwnorton.com with subject line “Norton Guide to AI-Aware Teaching.”

• If you want to read the e-book with a vibrant community of teachers this month, see the Perusall Engage book event below!

Events

• Norton is hosting a launch party for the book on Wednesday, July 8th, at 2:30pm ET. Marc, Derek, and I will be there to intro. the book, share our favorite parts, and have an “Ask the Authors” Q&A. You can register for the launch party for free here: https://us02web.zoom.us/webinar/register/9817823153827/WN_453PPw2WQfmYI9RuZdyBug

• Perusall is running an asynchronous Engage book club July 6th to August 2^nd. The e-book will be available on their social annotation platform for you to access for $15 for the session and an additional 4 weeks. Marc, Derek, and I have a reading plan and introductions to the material and will be engaging with readers throughout the month by commenting and responding in the text. Join other teachers to discuss strategies and share ideas! More info. and registration can be found here: https://www.perusall.com/events/norton-guide-to-ai-aware-teaching-engage . (We like Perusall and are happy for this publicity; however, just to be clear, we don’t have a paid sponsorship with them.)

Podcasts

Marc, Derek, and I have discovered that we like writing together, but we also like being interviewed together! We’ve done it a few times now. Check out these recent podcasts for our takeaways from the book.

• Social Learning Amplified (with video! hosted by Eric Mazur for Perusall Exchange), where we talk about our favorite examples from the book, approaches we admire, and challenges from AI.

• John Kane and Rebecca Mushtare’s Tea for Teaching podcast, where Derek reveals he’s never co-authored something before this, I liken our collaboration to a K-pop band, and we reveal why we put the chapter on academic integrity last.

• Derek’s Intentional Teaching podcast, where we offer takeaways on collaborating with colleagues, helping students find meaning in our courses, and how not to let AI get between us and our students.

• Bryan Alexander’s Future Trends Forum (with video!), a really lively conversation about how AI arrived in our classrooms and what we’re doing about it, plus great audience questions.

Table of Contents

Introduction: What Is AI-Aware Teaching?

Part One: Planning an AI-Aware Course

1. Evaluating Your Course Goals and Learning Outcomes

2. Meeting Students Where They Are

3. What Can AI Do?

4. Online and Hybrid Courses

5. Choosing the Role of AI in Your Course

6. Assessment Approaches

7. Drafting AI Policies and Guidance

Part Two: Teaching an AI-Aware Course: Assignments, Activities, and Assessments

8. In-Class Activities

9. Out-of-Class Assignments

10. Reading

11. Writing

12. Research

13. Using AI as a Tutor

14. Encouraging Academic Integrity

Conclusion: Working Together

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more.

Subscribe now

So, it depends.

This is why I’ve advocated for AI awareness: teachers knowing what AI is capable of so we can successfully steer our courses and students to good learning outcomes. Like a lot of technologies, AI can be useful and dangerous at the same time. The more teachers know about how to use it well, or to avoid it when it gets in the way of learning, the better. GenAI is out there and most students are using it, although they’re not necessarily using it to cheat and they have mixed feelings about it.

There are no easy responses to AI in education. In this newsletter, I’ve tried to be honest about that—while still providing resources and flexible approaches that address both AI’s benefits and its drawbacks. Many of you—from high school teachers to writing center directors to first-time teachers of composition—have reached out to me to thank me for that honesty and flexibility. I’m glad it’s been helpful! It’s a challenging as well as exciting time to be teaching in higher ed, and the journey is better with company.

So, I’m thrilled to announce that I’m now at work with Marc Watkins and Derek Bruff on a new book to support instructors, The Norton Guide to AI-Aware Teaching! You may already know Marc’s writing on AI for The Chronicle of Higher Education, his Substack, Rhetorica, or his work with the University of Mississippi AI Institute for Teachers. He’s a knowledgeable, thoughtful, and pragmatic voice on AI, writing, and higher ed, plus an award-winning teacher of online and hybrid courses. I cited Derek Bruff’s work on AI in workshops before I met him, and am grateful for the expertise he brings to this project: assessment, STEM education, ed tech, and beyond. This project will also rely on the keen editorial direction of Erica Wnek, who’s been behind this newsletter, many of W. W. Norton’s composition and rhetoric books including Everyone’s an Author, and more. She’s joining forces with Betsy Twitchell, Editor in Chief of Science at W. W. Norton, who oversees college textbooks for psychology, biology, chemistry, geology, and more. Marc, Derek, and I will be taking turns posting to this newsletter under its revised name AI & How We Teach, reflecting our expanded, collective expertise. If you’ve appreciated my voice in this space, you’ll love the additional guidance from these great writers and teachers!

How will The Norton Guide to AI-Aware Teaching be useful to you?

The Norton Guide to AI-Aware Teaching will provide practical strategies to navigate the uncertain and rapidly shifting terrain of AI, empowering college instructors to make choices about AI that are best for your courses and your students. In it, we’ll showcase tools, strategies, examples, and knowledge to help you redesign your course to be AI aware. Pragmatic and informed by research, we cut through the hype about AI to help you decide how you want to integrate AI in your course—or steer around it.

One of our early reviewers, Susan Ray, Associate Professor of English at Delaware County Community College, described the guide as:

a foundational, faculty-centered guide that cuts through the noise about AI in higher education, offering instructors both a conceptual framework and concrete entry points. It avoids the overwrought binary of boosters versus doomers (to quote Emily M. Bender) and instead gives us a shared language—embracing, resisting, and experimenting—for thinking through how AI fits into both our classrooms and disciplines.

Because we want this guide to be useful for instructors across the curriculum and across a spectrum of AI engagement, the guide offers:

• Practical Organization: Moving through planning, designing, and teaching an AI-aware course, the structure of the guide will help teachers implement new, effective strategies immediately.

• Flexible Guidance: Instructors can explore flexible, varied options for adapting their teaching to address AI’s impact on student habits and learning through frameworks that make AI-aware teaching both attainable and sustainable.

• A Focus on Learning Outcomes: Compelling research, thorough examples, and practical strategies empower instructors to make changes that best support their students.

• Useful Applications: Example assignments, activities, assessments, and course policies show concrete applications of AI-aware teaching.

• Student voices: The incorporation of students’ thoughts on how they’re using AI, how they wish they could use AI, and what concerns and delights them about AI can help to inform AI-aware teaching.

We want this guide to help build your confidence and agency with AI. We know teaching is most exciting when we keep students at the center!

You may wonder: why read a guide on AI when the terrain is so rapidly changing? How AI affects teaching and learning environments in the future will be impacted by new technological developments, lawsuits, regulations, social acceptance and resistance movements, supply chains, and international politics. This guide, which will go to press in Summer 2026, will inevitably reference technologies or capabilities that have shifted by the time you are reading it.

And yet, how learning works won’t change. The key principles of teaching and learning are well-established. You are a capable teacher, and you are teaching in a new situation. We know the general outlines of what AI can do and what it might be capable of in the future. So, this new situation doesn't have to be overwhelming. You can use the concepts of good teaching that you already know—or that you will learn through the guide—in order to respond effectively to this brave new world.

You may also wonder: how can I find the time to keep up with all this technological change and its impact on my courses? We are keenly aware how little time many of us have to respond thoughtfully to generative AI in our teaching. That’s why we’ve designed this guide with practical and actionable steps you can take now. Whether you are a graduate student teaching a course for the first time, a full professor with decades of teaching experience, or an adjunct faculty member teaching multiple online sections without much autonomy, the strategies and examples within The Norton Guide to AI-Aware Teaching are designed for you. Above all, we believe that the most impactful and powerful force in any learning environment isn’t a computer—it’s you, an engaged and informed educator.

On that note, we'd love to hear from you:

You can sign up to be notified when The Guide publishes here. The ebook will be available for free to W. W. Norton adopters, also available for purchase in ebook or print.

Thanks for reading AI & How We Teach! Subscribe and come back to hear more.

Subscribe now

1. How do I learn more about AI?

2. What do I do with it?

The best way to learn about AI is to use it. You can read about it and listen to podcasts, but there’s no substitute for trying out the platforms yourself—to see what they can do and what they can’t.

If you’re feeling stuck or unsure how to start, this post offers ideas for where to begin and how to level up. I also offer basics about some of the most popular platforms (as they currently exist) and suggest that you check out their privacy policies so you can decide how much you want to share as you try them out. (Bear in mind these notes are based on my personal experience and should not be taken as legal advice or analysis! And, of course, privacy policies and terms of use may change, so check them out for yourself to stay current.)

For a routine that can help you keep up to date, I recommend checking out Marc Watkins’s recent column for The Chronicle of Higher Education. On Substack, Eric Hudson offers a great list of resources to learn more about AI. And if you’re looking for an added challenge, Ethan Mollick’s current Quick Guide outlines research, voice, and video modes with major AI platforms.

Let’s dive in!

Which platforms should you start with? What are the risks?

The platforms I start with are ChatGPT (made by OpenAI), Claude (made by Anthropic), Gemini, and NotebookLM (both from Google). If your university has an institutional subscription to CoPilot (Microsoft) or another platform, then check those out as well.

All of these platforms currently have free access, although you’ll have to register with an email address. You may wonder: what does the company do with the information I put into the platform? What risks might I be taking? Here’s what the different companies say they do with your information:

• OpenAI (ChatGPT): ChatGPT collects your uploaded data by default—that is, your prompts and any writing or documents you share with them. According to their privacy policy, they may use that data to train future models. You can toggle a switch to turn off data collection by going into your account, then Settings, then Data, then turning off the setting that says “Improve the model for everyone.” That’s a nice bit of rhetoric, isn’t it? But actually, even if you opt out, it appears there are circumstances in which they can still use your chats to train models. So, read the policy carefully if you’re concerned, and avoid sharing anything that you don’t want reused in any way.

• Anthropic (Claude): Claude apparently does not collect your data by default. You can read more about their data privacy policies on their website. Anthropic seems relatively transparent about their data and models. (I noticed that they share Claude’s system prompts. A system prompt acts as a wrapper around any of your prompts, helping to steer the responses of a large language model like Claude.) Still, they say that they use customer prompts for research, and they may review your prompts if they trigger their legal guardrails.

• Google (Gemini): Gemini's privacy policy says the platform collects data for everyone over 18 by default. (Google is very familiar with the educational market for minors.) Their privacy policy describes the contexts in which human reviewers look at conversations, and they advise against entering confidential information into the app. Per the current policy, you can turn off “activity” by going into “Settings & help,” then Activity, and turning it off. Turning off activity can disable some functionality, however.

• Google (NotebookLM): Google says “we . . . never use your personal data to train NotebookLM.” They have the same caveats as Gemini about human review and being careful about what you share.

Not all platforms are the same, but given what's been reported about data scraping by AI companies, I am personally wary of anything I put into an AI application—regardless of what they say they do with my data. I don’t enter personal information, student work, or any unpublished work that’s not my own. I do enter my own writing into AI to experiment with it. (I also feel ok about entering material that's already on the web or that's open access, but you may have a different risk tolerance.)

Trying out AI platforms

If you’re comfortable sharing your own writing with the platform, then you might ask it to:

• Cut down the word count for a conference abstract.

• Suggest transitions between ideas, paragraphs, or sections.

• Make a written presentation more relatable to your intended audience, or make a handout with key points.

• Act like an editor for a target journal and give suggestions for how to pitch your piece, how to make it stronger, etc. Sharing the journal’s mission, editorial board, and an example article or website (if it’s open access) may improve its suggestions.

• Give you alt tags on images to improve accessibility.

Or if you’d like to try using AI to enhance your teaching, you could ask it to:

• Outline learning goals for a course, given the assignments, readings, and activities.

• Match learning goals from the course to each assignment (“Here are my learning goals, and here is an assignment from the course. Please tell me which learning goal the assignment contributes to and explain why. What additional skills is this assignment fostering?”).

• Consider alternative assignments that would meet the same learning goals of an assignment you share with the platform.

• Update your course syllabus with a new grading scheme such as labor-based grading or specs grading, etc.

• Read your syllabus like a first-year college student and let you know places to improve clarity.

• Give suggestions to make your course more accessible for students with executive dysfunction or ADHD.

• Design a rubric for an assignment you share.

• Generate visuals to illustrate your syllabus.

• Analyze data you collected from a pre-course survey to help you tailor your instruction to the class’s needs and experiences.

If you don’t want to share your own writing or course materials, you can still play around with AI and get ideas:

• Talk through your ideas for a new project. You can ask: “What are some key texts that I need to read to take on this project? What does the research on this topic look like already? In which fields would I find relevant work?”

• Ask the “Research” function of a platform (for example, Gemini, Claude, or ChatGPT) for a 5-page report on your topic. Skim it for ideas and check out the reference list for sources you may have missed. (Standalone platforms such as Elicit and Perplexity can do this, too.)

And here are a few ideas for using AI for fun or personal reasons:

• Make a meal plan for the week out of your CSA delivery, what’s on sale at the grocery store, or what’s closest to expiration in your fridge.

• Plan a potluck party for neighbors with a signup space for each category of items to bring.

• Help you think through gift ideas for someone who is hard to buy for.

• Generate a chore chart for your kids with tasks that are age-appropriate and possible incentives/accountability ideas to make it stick.

• Plan a vacation with some specifications: for X number of people, under a $ budget, with Y interests and Z amount of time.

If you’re looking to level up…

Here are some other platforms and AI interfaces you may want to try out:

NotebookLM: This platform from Google works a little differently from chat bots like ChatGPT, Claude, and Gemini. It was designed in collaboration with Steven Johnson, who has written over a dozen popular history books, which I think shows in its utility for writing. It’s also the tool that faculty I’ve worked with seem to find the most useful to design assignments around. The platform allows you to upload multiple documents in a “notebook,” where you can query them for summaries, particular information, etc. The “audio overview” feature produces a podcast using AI-generated voices talking about the topics in your notebook. Here are some ways I like to work with NotebookLM:

• Make a notebook out of one giant reference document I want to understand—and then query it or make a podcast.

• Collect some pdfs of a topic I'm working on and ask for key ideas, opportunities for further research, etc.

• Explore a topic I'm interested in but haven’t had time to dig into by uploading related documents and asking it to give me an audio overview or a plan for further learning.

• Upload some of my own writing or notes into it and ask it to characterize my style, help me write in new genres, or dig through relevant information.

• Use it as a reference finder after collecting my notes and quotes in a document; see Steven Johnson's website for good ideas on this use case.

Custom GPTs on ChatGPT: A custom GPT takes an OpenAI ChatGPT model and provides specific instructions and materials for it to draw on. While you need a paid account to create one, you can use public CustomGPTs with a free account. I’ve designed a custom GPT, AI Integration Bot, that gives the kind of feedback that colleagues interested in integrating into their course syllabi often ask me to provide. I loaded it with a bunch of my writings on AI, plus TextGenEd (which is open access and, I would guess, very likely in the training dataset for these platforms already), and tested it out on old composition syllabi of mine, before AI was a part of that course. It’s not the same as getting feedback from me personally, but I found that it offered some good suggestions for integrating AI activities into a course.

TextFX: From Google Labs, this little series of textual toys lets you find intersections between ideas, make a scene more surprising, offer lists of alliterative words, and more. It’s potentially useful for creative work, getting unstuck, or just some enjoyable wordplay! I find it to be very quick and easy to use, and the site includes short, instructional videos.

Thanks for reading, especially to subscribers! Watch this space for some exciting updates from Norton and me on AI and teaching support. I’ll be taking the rest of the summer slower here. When I’m not at the pool with my kids, you can find me on my personal Substack, Computation & Writing. And find more on AI and other teaching and learning topics on Norton’s Teaching and Learning Blog. Have fun this summer!

Subscribe now

It’s important for us to acknowledge our own feelings when students overuse AI, as Fasano does. We’ve put hard work into a course, maybe even overhauled it to be AI-aware, and yet here it comes, the AI paper: turned in at 11:58 p.m. with its “delving into” an issue using tripartite examples and smooth, even-handed arguments that seem disconnected to anything discussed in class. We can feel disappointed and even tell our students that.

And yet, our response to students can’t end there. As instructors, we have an obligation to support student learning, including how they make decisions about AI. So how do we talk with students when we think they’ve overused AI?

In this post, I offer a few strategies for opening up that difficult conversation, along with a perspective on academic integrity in the age of AI.

Why academic integrity matters

When we confer a degree, we’re certifying that a student completed a program designed to build knowledge and skills. Academic integrity policies protect the reputation and authority of the institution.

But academic integrity policies also help students. They help to guide student choices and communicate the expectations of an academic community. When a student joins our academic community, they have both an opportunity and an obligation to learn its practices. In turn, they should be able to depend on others in the community who are learning the same practices—they can expect the same from peers in their collaborative work, class discussions, and assessments.

Students get frustrated when their peers don’t meet community expectations. It feels unfair when a peer coasts on groupwork or skips out on assignments. Students in a focus group study we ran at Pitt this spring tended to think that using AI resulted in better grades. So, if a student is following the rules about AI, it feels unfair when their peers aren’t. Even students who understand the intrinsic value of what they’re doing can be annoyed when a peer gets a better grade by cutting corners.

And yet students aren’t in a position to call out a friend for overusing AI. It’s our job to pay attention. Not to police—but to notice, acknowledge, and start a conversation. Academic integrity policies are only a starting point.

Here’s where I should note that the work of paying attention can’t be outsourced to AI detectors. Detectors can flag phrases more likely to be produced by AI, but they only work on writing outputs: they can’t catch students using it to outsource the thinking, outlining, or reading for their writing. They produce false positives, especially for non-native English speakers. But worse is what Janae Cohn notes from her perspective as executive director of the Center for Teaching and Learning at U.C. Berkeley: “While we recognize that A.I. detection may give some instructors peace of mind, we’ve noticed that overreliance on technology can damage a student-and-instructor relationship more than it can help it” (as quoted in the New York Times). AI detectors cannot save us.

My colleague Jeff Aziz, who is extremely sharp on AI and who serves as our Academic Integrity Dean at Pitt, says: “A first academic integrity violation is an opportunity for a conversation.” He encounters students who felt desperate and made bad choices. He says he hasn’t yet caught a student trying to skip through the entire course of study using AI; instead, he sees students “experimenting” with the boundaries of academic integrity. Students enrolled at Pitt are in a community of people responsible to each other—which is why academic integrity is important—and sometimes they need to learn that, he says. After his conversations, Aziz sees very few second offenses.

In the classroom: an opportunity for conversation

We’re all familiar with the “teaching moment,” when something unexpected happens and we turn it around so students can learn from the situation. So, when you get that AI-suspicious paper, let your conversation be a moment for teaching rather than punishment. Rebecca Moore Howard, who argued against the “academic death penalty” for plagiarism advised: “often a pedagogical rather than judicial response is appropriate.” We must separate our affective reactions (disappointment, anger, betrayal) from our pedagogical responses.

When you receive the AI paper, first ask yourself why you think the student used AI on this assignment. Did the writing not match what you expected? If so, why do you think so? What if you’re not sure whether they overused AI or not? It’s never a bad thing to talk to students about their writing, so even if you're not sure, it's okay to reach out to them. By framing your outreach as conversation rather than accusation, you can use it as an opportunity to teach them about writing.

In your conversation,

• Ask the student to expand on the ideas they wrote about: “tell me more about…” or “what were you thinking when you wrote this?" Whatever they turn in to you, they are responsible for their words. If they are unable to account for their own work, then they aren’t fulfilling what you’ve asked.

• Ask them to tell you about their writing process. Are they taking notes by hand? Are they keeping track of sources in a separate document? Are they seeking feedback from others, including friends, writing tutors, and AI? You can ask them directly: “what technologies did you use in composing this? Did you use AI at any stage of your writing?” Try to figure out if they see themselves as collaborating with AI, or if they’re looking for direct outputs (see Anthropic’s study of student uses of Claude for more on the difference).

• Probe the context of the student’s choice. In my research, I’ve seen that students overuse AI when they’re pressed for time, when they don’t find work meaningful, and when they’re in over their head. Rather than asking a student to explain their choice, you can begin by asking the context of their work in your class: how are they scheduling time to complete it? What are their interests and goals, and how do they see work for your class related to those goals? Are they finding your course too challenging? (See The Opposite of Cheating: Teaching for Integrity in the Age of AI for more contexts and responses.)

• Remind them of your policies and those of the university. Make sure they understand what's allowed in your course. (Ideally, students are involved in making these policies, as Sarah Elain Eaton argues.)

• Remember that it's impossible to prove a student has used AI in their writing unless they tell you they have done so. AI detectors aren't reliable, ChatGPT won’t tell you truthfully if it wrote a paragraph, and you can't recreate the original text as you can, say, if a student copies text from a website.

• If you think a student has used AI in a way that's inconsistent with your AI policy, consider letting the student retry the assignment, revise what they've turned in, or otherwise find a way for the student to learn from the assignment as well as get some credit for their work. Reach out to colleagues, department chairs, your Center for Teaching and Learning, or Academic Integrity officials for support.

And then sometimes, it just doesn’t work. You spent two weeks describing the AI policy, showed examples of possible collaborations, asked for disclosure statements, and the student still turned in a stiff paper but refuses to admit that it’s AI or talk openly when you ask them about it. The student just isn’t ready for the conversation or the learning.

One teacher online lamented that after he failed a student for an AI-written research paper, she emailed him back with an AI-written apology. He received a lot of suggestions for his next move, but my favorite was a prompt for ChatGPT: “Write an email, from YOURSELF apologising to a student that she failed her research project because she used chatGPT to produce it.” Follow the link to try the full prompt yourself! The results don't disappoint.

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more.

Subscribe now

Student data tells us a more complicated story. For example, from my own study team's work talking to students across four of the University of Pittsburgh campuses, I learned that: (1) students often think they know about AI because they’re users of AI; (2) students need more examples of what productive AI use could be because most of them are finding uses on their own—not all of which seem great; (3) most faculty have no idea how prevalent, varied, and complex AI use is among students, and students sense that gap.

So, how prevalent, varied, and complex is AI use among students? Recent reports from AI companies, polling companies, and educational companies can help us to understand. In this post, I’ll give an overview pulling from 10 reports that together provide a good snapshot of this moment. (You’ll find a list of links and brief descriptions of the reports at the end of this post.) Read this post once to get the gist, and then keep it bookmarked for arguments you might need to make to support faculty training, new course development, or respond to administrative pressures. Data matters!

A note about the sources: with the exception of one study (HEPI), the reports I’m drawing from are focused on the United States—the biggest educational market. It’s also worth noting that two reports are from AI companies (OpenAI and Anthropic) who have recently announced new products and partnerships with American educational institutions. These AI companies are actively tipping the scales to encourage student use through partnerships and promotions (ChatGPT Plus is free for students through May 2025, and Gemini Advanced is free for college students through “finals 2026”). The reports from OpenAI and Anthropic both end by emphasizing the importance of teaching students to use AI responsibly, and OpenAI’s report is particularly focused on promoting government and educational sponsorship of AI literacy programs for students so that they can prepare for the workforce. So, you know, grains of salt.

How prevalent is AI use among students?

Studies vary widely on their measures of student AI use, but it’s highly prevalent and increasing dramatically. And we’re probably not seeing the extent of it: only 30% of students thought their instructors were fully aware of all of their uses of AI (Educause) and 54% of higher ed leaders felt their faculty were good at detecting students’ AI use (AAC&U). These stats help paint the picture:

• 79% of Gen Z use GenAI, and 47% do so on a weekly basis (Walton-GSV-Gallup). Teens 13–17 doubled their adoption from 2023 to 2024, from 13% to 26% (Pew Research). The UK group HEPI also reported a surge in student use over a yearlong period: from 66% in 2024 to 92% in 2025.

• Students are increasing their use for assessments: from 53% in 2024 to 88% in 2025. Use of AI to generate text has doubled from 30% in 2024 to 64% in 2025, and almost one-fifth of students report using AI-generated text directly (HEPI).

• Surprisingly, 43% of students reported not using AI (Educause). This number is high, and Educause suggests students might have underreported their AI use. My most recent survey data (December 2024) indicates that 32% of students never use AI.

• Very few students totally resist AI—most students in my focus group study were using it at least in some capacity. If I had to estimate use based on the focus groups, I’d guess >90% of students are using AI.

• AI use among students is influenced by instructor policy, although policies aren’t fail-safe. Whereas 72% of students reported not using AI if it was prohibited, only 6% of students didn’t use it if it was allowed (Educause). My survey data from Pitt students indicates that while 59% of students used AI when instructors didn’t specify a policy, 12% of students still used it when it was disallowed.

The conclusion I draw from all of this is that there is widespread use of generative AI among undergraduate students and considerable variability in reporting of that use. Almost all students have tried it. Some students are holdouts, or they may use it so infrequently they say they don’t use it on a survey.

What are students using it for, and why?

Students turn to AI to save time, boost their grades, and get help quickly (HEPI, GenAI Conversations) as well as to avoid work they find less meaningful (GenAI Conversations). Some of the students we talked to in GenAI Conversations said they were more likely to use it when a deadline loomed. Others said they used it for feedback because it wouldn’t judge them. HEPI also cites personalized support, improving AI skills, and improving learning as factors for students choosing to use AI. Half of students felt that using AI improved the quality of their work (HEPI). Our study team also heard from students who felt AI improved their grades, so they felt it was necessary for their success and to compete with peers. Their responses made us think about how we might reframe office hours or Writing Center consultations, and the role of grades and deadlines in our classes.

Students use AI for every aspect of the writing process: generating ideas, aiding research, summarizing readings, drafting text, revising, and editing. Educause notes that common uses include brainstorming (33%), refining ideas (24%), and organizing (24%). Top uses of ChatGPT include starting papers, brainstorming, summarizing texts, and editing writing. Claude has similar uses among students: essay editing, generating practice questions, summarizing of academic texts. Anthropic notes that students also use Claude to provide technical explanations or solutions for academic assignments, including debugging code and solving math problems, and translating or proofreading content between languages.

Anthropic’s report uses Bloom’s taxonomy to categorize student use and indicates that students are using AI for higher-order cognitive tasks, namely creating and analyzing—although much less so for evaluating—another higher-order task. Students used it about half of the time for what Anthropic calls “Direct answers,” and half of the time for “Collaborative conversation.” Direct answers may suggest academic integrity concerns, and Anthropic gives a few examples they were concerned about: providing answers to multiple-choice tests, rewriting texts to avoid plagiarism, and providing answers to English language test questions. Collaborative conversations might include asking for feedback or explanations for concepts.

Since Anthropic’s analysis indicates higher-order cognitive tasks, we may worry that students are offloading these tasks to AI, rather than developing associated skills.

As a teacher, I feel more comfortable with students using AI in collaborative ways rather than direct answers—although I worry about them using AI to substitute for feedback from peers, tutors, and instructors.

Are there differences across fields and demographics?

Use in the humanities is less than in the sciences and business (HEPI, Anthropic), perhaps due to attention to authorship and originality of expression in the humanities and a greater stigma associated with AI use. And while some students perceive AI as a boost for their grades, they see the effect as lower in their humanities work (HEPI). Computer science students appear to be the greatest users (Anthropic: 38.6% of conversations, with only 5.4% of US bachelor’s degrees), with Humanities among the least (Anthropic: 6.4% of conversations, 12.5% of US bachelor’s degrees).

Women are adopting AI at a lower rate than men. There appears to be a greater fear of cheating among women, and men tend to adopt cutting-edge tech at a faster rate than women (Walton-GSV-Gallup). A 2023 BestColleges survey found 64% of men reported using AI tools for coursework versus 48% of women.

There are some statistics on adoption across race and income. According to Pew, Black and Hispanic teens 13–17 were more likely than White teens to have used ChatGPT for their schoolwork in 2024 (31% versus 22%). These demographics were even in 2023, and all of them were lower (11% each). At the same time, White teens (83%) were more likely than Black (73%) and Hispanic teens (74%) to say they’ve heard about ChatGPT. The Gen Z survey reported slightly different results: Asian and Black Gen Zers were more likely than White or Hispanic Gen Zers to use GenAI frequently. Income influences knowledge of AI: teens from households with higher income are more likely to have heard of ChatGPT: 84% of teens in households making $75,000 or more, versus 67% of teens in households making under $30,000 (Pew).

The AAC&U report on higher ed administrators (January 2025) indicates that AI use among faculty trails behind that of students—and that of administrators. Tyton Partners reported the same phenomenon in Spring 2024: most students were familiar with AI or used it regularly, whereas 50% of advisors and other frontline staff had never used AI. Administrators and instructors who had never used AI were more likely to think it would have a negative impact on student learning than those who used AI regularly. That administrators tend to use AI more than faculty might explain the July 2024 findings from The Chronicle of Higher Education: administrators are more excited than faculty about AI integration in education.

What concerns students about their AI use?

Students have mixed feelings about GenAI. Many are anxious about potentially losing critical analytical skills (49%, Walton-GSV-Gallup). More Gen Zers feel only anxious or angry about AI (30%) than feel only excitement or hope (26%) and 17% say that they feel both positive and negative emotions about AI. The Gen Z survey also indicated strong preferences for human doctors, drivers, and tutors—countering some claims that humans won’t be needed for those jobs in the future. So, Gen Z may be using AI frequently, but they are not all-in.

Students see a mismatch in their perceived future with GenAI and the training their institutions are providing. The majority of students in the Educause report (55%) expected to use generative AI in their future career, but only 20% of them felt they were getting appropriate training from their institution. Students who felt their institution had a cutting-edge approach to technology, including AI, felt more prepared for their career than students who felt their institution was behind the times (Educause).

In the classroom

So what does all this mean for our work as teachers? What actions can or should we take as one term winds down and the next looms ahead? Here are four areas I’ve been reflecting on as I plan for the term ahead.

1. It’s important to acknowledge that AI is everywhere for our students, and the majority of them are already using it in their coursework. Even if they don’t use it, their friends do. Students in our focus groups indicated that even if they were opposed to AI, or if there were course policies against it, they used it “just a little bit.” They are constantly making choices about AI use. Sometimes they can’t choose though: their apps and educational software foist AI on them, which students called “unnecessary.”

2. Students benefit from open conversations about AI in our classes. Their uses are so varied—sometimes productive and sometimes creative—but not always what we’d call beneficial to their learning. Students need concrete examples of what constructive AI use looks like for specific assignments (see my previous post on AI-Aware Teaching Examples).

3. We need to offer transparent assignments and explain why the work should matter to students. Work that students find to be “busywork” is a candidate for AI. If an assignment isn’t clearly connected to student learning outcomes, it should go. If it is connected, students need to understand exactly how. Channeling Marie Kondo, I’ve been looking closely at my assignments and asking myself which are likely to spark meaning—and even joy—for my students.

4. We should make students feel more welcome. Students sometimes choose AI in contexts where human feedback or their own work might benefit them more. They turn to AI for affective reasons: to avoid being judged, to improve their perceived self-efficacy. They tend to avoid office hours. How can we truly invite them in? Or, for students who turn to AI to meet deadlines or make better grades, would flexible deadlines and alternative grading schemes help to refocus students on achieving learning outcomes?

Sorting through all this data, I think about the public service announcement, the more you know. The more we know about how they’re using AI, the better we can do for students learning now.

Subscribe now

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more.

Studies cited

• AAC&U. (2025, January) Leading through Disruption. https://www.aacu.org/research/leading-through-disruption. NOTE: This report focused on higher education leaders, 334 of whom responded to the survey. Institutions were diverse in terms of student population size.

• Anthropic. (2024, October 26). Anthropic Education Report: How University Students Use Claude. https://www.anthropic.com/news/anthropic-education-report-how-university-students-use-claude. NOTE: Based on one million anonymized student conversations on Claude.ai; they relied on .edu addresses, so they could also be capturing staff and faculty conversations, plus they didn’t capture conversations that occurred on non .edu addresses.

• BestColleges. (2023, November 22). 56% of College Students Have Used AI on Assignments or Exams. https://www.bestcolleges.com/research/most-college-students-have-used-ai-survey/. NOTE: A 2023 survey, which included 1,000 undergraduate and graduate students.

• The Chronicle of Higher Education. (2024, July). How Generative AI Is Changing the Classroom. https://www.chronicle.com/featured/digital-higher-ed/how-generative-ai-is-changing-the-classroom. NOTE: This report is based on faculty interviews (including me!) and surveys of higher ed administrators and faculty.

• Educause. (2025, April 14). 2025 Students and Technology Report. https://www.educause.edu/content/2025/students-and-technology-report. NOTE: The report has a section on generative AI. They surveyed 6,468 respondents from 37 higher education institutions.

• Gallup & Walton Family Foundation. (2025, April 8). Voices of Gen Z: How American Youth View and Use Artificial Intelligence. https://www.gallup.com/analytics/651674/gen-z-research.aspx & https://www.waltonfamilyfoundation.org/about-us/newsroom/gen-z-is-using-ai-but-reports-gaps-in-school-and-workplace-support. NOTE: The report samples 3,465 13–28 year olds living in the United States. Their emphasis is on Gen Z, and not students per se.

• HEPI. (2025, February). HEPI / Kortext Student Generative AI Survey 2025. https://www.hepi.ac.uk/wp-content/uploads/2025/02/HEPI-Kortext-Student-Generative-AI-Survey-2025.pdf. NOTE: The report is based on responses from 1,041 undergraduate students in the United Kingdom.

• OpenAI. (2025, February 20). College Students and ChatGPT Adoption in the US. https://openai.com/global-affairs/college-students-and-chatgpt/. NOTE: The report includes data from a survey (n=1,200) and user data of US-based 18–24-year-olds. It’s unclear to me how they know they’re capturing student data.

• Pew Research Center. (2025, January 15). About a Quarter of U.S. Teens Have Used ChatGPT for Schoolwork—Double the Share in 2023. https://www.pewresearch.org/short-reads/2025/01/15/about-a-quarter-of-us-teens-have-used-chatgpt-for-schoolwork-double-the-share-in-2023/. NOTE: A report on teens 13–17 from Fall 2024, frustratingly, their survey only asks about ChatGPT, which is an imperfect proxy for AI use generally.

• Tyton Partners. (n.d.). Listening to Learners 2024. https://tytonpartners.com/listening-to-learners-2024/. NOTE: They surveyed 1,600 students across 850 institutions, with 26% being first generation students.

• University of Pittsburgh. (2025, March, unpublished). GenAI Conversations. NOTE: This is my own study, collaborating with colleagues across four Pitt campuses, citing focus groups with 95 students total.

• University of Pittsburgh. (2024, December, unpublished). Student Uses of Writing Technologies Survey. NOTE: This is my own survey study of Pitt Composition students (n=167).

I was wrestling with this question as a writing program administrator in early 2023, when I received a gift from an unlikely source: the preeminent science journal, Nature. Just a couple of months after the launch of ChatGPT, Nature came out with a policy disallowing authorship credit for AI:

[N]o LLM tool will be accepted as a credited author on a research paper. That is because any attribution of authorship carries with it accountability for the work, and AI tools cannot take such responsibility.

I couldn’t believe it: embedded in this AI policy is a theory of authorship and accountability! I seized on this theory when crafting our policy at Pitt.

I wanted a policy for our Composition program that would accommodate instructors who embraced AI use in their courses as well as those who resisted it. Nature’s idea of responsibility and writing fit with both the constraints of our academic integrity policy and the diversity of approaches among our instructors. So, I included this in our AI policy: “You are the author of your work for the course and authorship means you take responsibility for your words and claims, regardless of which tools you use.” (See the full policy at the end of this post.)

The challenges of AI for writing classes aren’t resolved by policy alone, so—I hate to say it—there is really no right AI policy. Instead, there are many ways to write an AI policy for a course or program—as demonstrated by Lance Eaton’s exhaustive online collection, which includes 186 policies! (Helpfully, Eaton provides an overview of the policies on his Substack.) Among all these choices and difficult balances, this post will help you to create thoughtful guidance for students—the first step toward helping them make good choices about AI.

Why write an AI policy for your course?

Students are looking to instructors for clarity. Inside Higher Ed reported data from May 2024 indicating that only a third of students knew when or how they could use AI for coursework. And the 31% of students who did know about appropriate uses of AI credited faculty for their knowledge, not university-wide policies.

In my survey of Pitt composition students, which I’ve administered every semester since December 2022, students overwhelmingly ask for specifics on how they’re allowed to use AI in their courses (>90%). In response to my open-ended question, “What would you like to tell your instructors at Pitt about any of these [AI] writing technologies or your use of them?,” many students asked for clearer guidelines, saying things like:

“Make clear guidelines about usage.”

“Create an outline of how you want them to be used.”

“I think it is okay to allow students to use these programs in class, but a class policy should be outlined in the syllabus on how they should be used.”

“I want to know if it counts as cheating to use them.”

Many students want to follow the rules of a course. But generative AI is so new, university policies are so broad, and students receive so many mixed messages (think: AI marketing, their friends, a bunch of different instructors) that it’s hard for them to know what’s right for each course without explicit guidance.

In a recent opinion piece for Inside Higher Ed, Dan Cryer, an Associate Professor at Johnson County Community College in Kansas, describes the uneven terrain a student must now navigate as they consider using (or not using) generative AI. Even when we think we’re being clear, we’re often not specific enough. Many policies require an “AI acknowledgment statement,” for instance, but just how much should a student disclose? If an AI platform spills more help than the student asked for, should they try their best to ignore it? Cryer calls this “responsibilization.” Universities are shifting the responsibility of maintaining academic integrity, once shared between teachers and students, down to students alone. Students are bewildered, he argues, and it’s not fair to ask them to navigate the ethical challenges of generative AI without good guidance.

While I think we should emphasize students’ responsibility for their work and their role in maintaining academic integrity, I agree with Cryer that it’s our job to give them guidance. Cryer suggests two ways instructors can alleviate the responsibility burden that AI puts on students: (1) talk to them empathetically about this challenge, and (2) provide spaces to write that are free from AI, which gives them a break from their relentless decision-making about AI. (My previous post on AI-aware courses offers more strategies for helping students make good choices about AI.) Being clear about what we expect in our courses is the least we can do to support students newly “responsibilized” by the challenges generative AI presents. We also need to discuss our expectations for their use of AI.

Connect your AI policy to course learning goals

We like to think of our writing courses as important: they teach crucial skills and even inspire students! And they are important! But it’s also true that students are overburdened and struggling with balancing their time and commitments. A writing gen-ed might not always be at the top of their list. As they’re making decisions about AI use, how do they balance their time, attention, and need for attaining skills?

A policy on a syllabus is a good starting tool—but it isn’t enough. Syllabi have become pseudo-contracts: a lot of fine print that describes terms, but few students actually read them. To help students understand what a policy means, we need to talk about it with them—offer examples of acceptable and unacceptable uses of AI and explain why. Students responding to my survey often express gratitude for these conversations. One wrote, “I appreciate that many instructors are willing to have productive conversations with their students about AI in the classroom."

Students are always making choices about whether and how to use AI. Cryer reminds us that, ultimately, it’s up to them. And yet we can and should support their agency. For example, Tim Laquintano and I wrote a letter to students, explaining some of the workings, benefits, and drawbacks of generative AI. Jennifer Sano-Franchini, an Associate Professor at West Virginia University and an advocate for refusing AI, offers her students a detailed background on generative AI’s ethical and functional shortcomings. Ultimately, students are more likely to make good choices about AI if they see their work connected to course outcomes and their own learning goals.

In the classroom

To develop or refine your own policy, consider what’s most important to you and your course, and how you might use a policy to emphasize those values. Since my courses generally allow for some uses of AI, I value flexibility in process and emphasize students’ ultimate responsibility for the work. What are your key values?

Here are some other tips as you hone your generative AI policy and implementation:

• Think about how your assignment design works alongside your policy to help students make good choices about their uses of AI in your course.

• Discuss your policy with students at the beginning of the term. When you introduce a new assignment, discuss how the policy applies to it.

• Consider including students in your policy design or have them come up with examples of acceptable use for specific assignments.

• Offer avenues for students to ask questions about your AI policy.

• Work with colleagues or your teaching center to harmonize policies as much as possible.

Below are some examples of AI policies I find helpful. Feel free to share your policies or policy considerations in the comments!

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more.

Subscribe now

Some example AI policies

My AI policy (used in composition classes at University of Pittsburgh):

The use of generative AI writing tools (such as ChatGPT, Claude, Grammarly, Gemini, Copilot, or others) is allowed in this class within specific contexts and only if such use is properly acknowledged. Assignments for the course have been designed to help you develop as a writer, and some of them may call on you to practice writing with the help of such tools. As your instructor, I will assume that any use of these tools will be only within the contexts the assignment allows (for instance, you can use ChatGPT for brainstorming if the assignment asks you to do so). You must acknowledge the use of AI in your assignment in an "Acknowledgement of AI Use" statement that:

-Specifies which technology was used and on what date (ChatGPT, Claude, etc.)

-Identifies the prompts used or includes links to your chats

-Explains how the output was used in your work

The use of AI outside of contexts where the instructor specifies its use, or failure to acknowledge any use of AI technologies in your work will be considered an academic integrity violation and addressed according to Pitt’s Academic Integrity policies. You are the author of your work for the course and authorship means you take responsibility for your words and claims, regardless of which tools you use. Please see me if you have any questions about this policy.

Bryan Hanks, in the Anthropology Department at Pitt, offers specific guidance on AI for weekly assignments listed on his syllabus for “Peoples of the North.” To augment these specific policies, he goes over acceptable use in class. A couple of example weeks are below:

Week 3 AI GUIDELINES: You may use AI to help explain complex climate science concepts from the readings. Practice writing prompts that extract specific information rather than general summaries. Document all AI prompts that you use and your own analysis in your weekly notes.

Week 5 AI GUIDELINES: You may use AI to explore these concepts (or those you choose), the types of datasets that may be available for analysis, and academic resources that would be useful in developing your story map. Document all AI prompts that you use and your own analysis in your weekly notes.

The Refusing Generative AI in Writing Studies Guide offers detailed guidance for developing policies that help students choose not to use AI. One example is the policy from University of Arkansas’s Program in Composition and Rhetoric:

Using ChatGPT for Your Work in This Class: This is a course that asks you to work through the writing process and develop your own strategies for how to tackle each step of the process, from idea generation to drafting to revision to reflection. For this reason, I ask that you do not use artificial chatbots like ChatGPT to complete any portion of the assignments for this class.

My Pitt English colleague Sarah Hammock limits use of AI and emphasizes how she values students’ voices in her Science Fiction class’s AI policy. She notes that her students have responded well to this policy:

AI is rapidly becoming nearly ubiquitous in reading and writing. In this class, alongside our other work, we will consider when AI tools will be helpful, how to use AI tools effectively, and when using AI might negatively impact your learning or performance.

We will also discuss some of the critical issues that AI raises: How do you want to use AI in your work and daily life? How might AI affect the society we are building for ourselves and our descendants? How can science fiction help us imagine and guide the ethical and human impacts of AI technology?

I will permit or encourage you to make use of AI tools when I believe it could help you practice useful technology skills or save you time and energy.

In turn, believe me when I say that I prefer to read and interact with text and ideas written in your own imperfect human voice, without polishing, cleanup, clarification, or restatement from Grammarly, ChatGPT, or any other AI assistant – and seeking aid from these sources will not improve your grade in this class.

The overall position of this course is that you are a reader, writer, and thinker, and AI is a tool you may sometimes use to supplement – but never replace – your reading, writing, and thinking.

In this post, I drill down to some examples of AI-aware versions of common writing assignments: annotated bibliographies, reading responses, and reflections. By using pathways that acknowledge AI, we can achieve learning goals that support students’ writing, reading, and thinking development. And we can do it without resorting to unreliable AI detectors.

Stress testing assignments with AI

Prior to using assignments with students, think about stress-testing them with AI. What are the vulnerabilities in your assignments that AI now exposes? Some previously great assignments are now easy for AI to complete, which then presents students with a temptation to shortcut learning. The goal is to set up an assignment that, by default, leads students to make good choices about AI use.

To stress-test your assignment, first break down the tasks required for completing the assignment. Then ask: What are the goals for each task in the assignment? Which steps are most vulnerable to AI? Test this by prompting AI to do one or more of the steps. Which parts of an assignment do you want students to do on their own, without AI? Which parts can or should they use AI for? You can see Tim Laquintano stress-test and redesign an assignment he’s used at Lafayette College in a resource he and I put together for the University of Pittsburgh Writing Institute.

Let’s look at a few common writing assignments and how we might approach their redesign by either resisting or embracing AI.

In the classroom

Once you break down each assignment into its component tasks and goals, you can consider how you might approach an AI-aware redesign by resisting or embracing AI.

Annotated bibliographies

Writing classes have often used annotated bibliographies as a scaffolding step to help students prepare for research. However, AI is very good at this task. I tried the following prompt with both ChatGPT and Claude, and received a decent draft of an annotated bibliography. Each tool provided different sources, so if I were a student wanting to shortcut this assignment using AI, I could combine the results, rewrite them lightly, and turn in my assignment. Try it yourself in an AI platform of your choice:

Please give me an annotated bibliography on the pros and cons of traffic calming measures on college campuses. I would like 4, high-quality sources, cited in APA format, and about 5 sentences explaining the argument of each source and how the source contributes to a potential research paper on the topic.

How can we revise an annotated bibliography assignment to help students make good choices about using AI?

First: what is the goal of having students do an annotated bibliography? I might say: to practice finding and evaluating sources; to get familiar with campus library resources; to prepare for a project that draws on sources to make an argument; to discover academic approaches to personal interests; etc. Now: How can we meet those same goals while acknowledging that AI does this task well?

• Here's a potential resisting AI approach to meeting the goals of a traditional annotated bibliography: Have students write an “I-Search” narrative about their research process, describing why they chose a topic, how they began their search, who they talked to, what threads they followed and why, what questions remain for them, etc. The I-Search approach, introduced by Ken Macrorie long before AI, emphasizes personal discovery, student agency, and writing process over research products and argument. Those are moves students will benefit from now, too. Stacey Waite at University of Nebraska has students prepare for research by making a list of what they don’t know about a topic, what they want to know, and what they can’t know about a topic. You could ask students to begin that list in class, minimizing distractions and temptations of AI. Elaine Minamide, an instructor at Palomar College, describes her approach and offers example student I-Search papers. She notes that the I-Search activity “encourages discovery as opposed to conclusion” and helps students find their own voice and style by narrating their experiences. ReadWriteThink offers a helpful strategy guide on I-Search that’s geared towards secondary education, but easily adapted to higher ed.

• An embracing AI approach: Have students go to a site like Elicit or SciSpace and ask their research question. These sites allow users to filter for “top quality” journals, provide abstracts, etc. Both sites connect to the Semantic Scholar database, which provides links to real published research papers, so the chances of hallucinations are minimal. By showing students how to evaluate sources by number of citations, quality of journal, and relevance of the abstract to their research question, you can help them learn good research skills at the same time that they pick up some skills in working with AI.

Reading responses

We often have students do low-stakes writing to prepare for larger, more complex assignments. One example is reading response posts where students respond to a question about an assigned text via an LMS discussion board, like Canvas. I’ve done this for both grad and undergrad classes. But, as one student recently told me, these discussion posts are now “AI central.” Ouch.

What’s the goal? Often the point of reading responses is to hold students accountable for doing their assigned reading, have students interact with each other, and motivate prep for a discussion in class—or occasionally serve as a substitute for in-person discussion.

• A resisting AI approach: You could eliminate this kind of assignment altogether and shift the work to in-class discussions. Or, you could have students print out a reading and hand-write their questions and responses. Platforms like Perusall are good for doing this digitally. Jason Crider, at Texas A&M, offers a “trick” assignment that would work in this context; he calls it the “paranoid memorandum.” First, have students break into groups to work on a brief writing assignment. Each group is given an instruction to either use AI or not use AI in completing the assignment. Groups then share their work with the entire class and guess whether each group used AI or not. The trick is that all of the groups are assigned to not use AI. The discussion, Crider explains, can focus on what it means to be readers in a context where anything is potentially AI. This approach will only work once, of course!

• An embracing AI approach: You can have students interact with a text using AI before assigning it. First, give students a heads-up that they’ll be working with a particular text in class, but tell them not to read it yet. In class, give them 10 minutes to ask an AI of their choice anything they want to know about the text: a summary, which texts or debates it engages with, how it connects to concepts in class, etc. Claude, ChatGPT, or NotebookLM work well for this task. Then, run a discussion about what they learned about the text, and what further questions they have. Finally, you can assign students to read the text themselves prior to the next class. Primed with these questions and discussion, they’ll be more prepared to understand a complex text. My Pitt colleague Matt Burton came up with this idea, which worked great in our co-taught Writing Machines class.

Reflections

Some advice on discouraging students from using AI focuses on having students make personal connections in their writing. But some teachers report that students are even using AI for their reflections—offloading the discussion of their thoughts and writing development to AI. Unequivocally, I’d say: this is not an appropriate use of AI. How can we prevent this from happening in our classes?

• A resisting AI approach: In class, have students annotate their own work or do pen-and-paper brainstorming: how did they feel when writing this assignment? What challenges did they encounter? What did they learn? Or, have students interview each other about their approach to the assignment. They can develop questions for each other or you can develop these questions as a group, guiding students through the metacognitive work of how to understand their own thinking. They can then interview each other in class, which incidentally means they (1) get to know each other, fostering class community; (2) learn from each other about different writing processes and ideas; and (3) have some accountability to each other, not just to you. (Imagine a student improvising their way through an interview with a peer when they’ve used AI to complete their assignment—awkward!)

• An embracing AI approach: You could have students do a human vs. machine contest, where they try to make the most human reflection using AI, then test it out on each other. Which part was done by AI? How can they tell? What did they learn from this exercise and what didn’t they learn?

For all these approaches, especially those resisting AI, there’s the challenge of having limited time: you can’t have them do all the work in class! Consider the balance of where AI might be helpful, and how students can learn to interact with it in ways that support their learning.

P.S. If you’d like to see an advanced example of stress testing a common genre of writing, check out Arizona State University professor Andrew Maynard’s recent post, where he used OpenAI’s new Deep Research tool to write a PhD Dissertation! (Spoiler: Andrew put in a lot of work, and the result was just OK and potentially passable.)

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more ideas for AI-aware approaches to teaching writing.

Subscribe now

"I think students should not [learn about AI in their courses], however, in order for that to happen the structure of classes must change to not allow for AI to be used. Using AI is addictive so honestly if I can use it I will."

It’s that last sentence that gets me. This isn’t the only time I’ve seen a student talk about “addiction” to AI. I read this as a cry for help: help me to avoid self-sabotaging uses of AI, because without a course structure that discourages me, I’m likely to make bad choices.

So, how can we help students make smart choices about AI in their writing?

AI is changing writing, and the worst thing we can do now is pretend that’s not true. Writers encounter AI in increasingly sophisticated autocompletion, app integrations, and their word processing programs. At this point, corporate writing is 10–24% AI-assisted. Students see ads from Grammarly telling them they can “win” at writing and influencers peddling AI “homework help.” ChatGPT even got a Superbowl ad this year.

With AI at our fingertips and in our faces, the choice of opting out of it while doing any form of digital composition is getting increasingly difficult. The default is AI, and settings to turn it off, if they even exist, are well hidden. AI’s incursion into our writing environments and the efficiency rhetoric that accompanies it is almost completely out of our control, and our students are struggling to make good decisions about their use of AI. The student who writes that “using AI is addictive” is, I think, asking for a respite from the relentlessness of AI in their life.

What can a writing teacher or program do?

AI-aware curriculum

We may not be able to control the degree of AI’s encroachment in writing environments our students use, but we can control how we structure our own course’s engagement with AI.

A writing curriculum should address AI skills, and first-year composition (FYC) should promote some critical AI literacy, but not every writing class needs to integrate AI. However, all classes—even those outside of writing programs—should acknowledge that AI is now out there. That means revising writing assignments that no longer lead to desired learning outcomes. If AI completes an assignment too easily, some students may be tempted to use it to shortcut learning goals—even if they know it’s a bad choice.

How we integrate AI into our writing courses should follow from our curricular and course goals. What do we think students should learn in our courses? What should our undergraduate majors know and be able to do when they graduate? We must tune our courses to meet those goals, acknowledging AI as part of our shared writing landscape, and providing structure in which students can make informed choices. I think of this approach to curriculum and course design as AI aware.

For some courses, being AI aware means actively refusing AI. The Refusing GenAI in Writing Studies guide by Jennifer Sano-Franchini, Megan McIntyre, and Maggie Fernandes, provides a framework for “individuals and/or groups [to] consciously and intentionally choose to refuse GenAI use.” They outline principles about linguistic homogenization, relationships between language and power, ideological bias in technologies, labor issues, and environmental impacts of AI as reasons to refuse AI, arguing that “refusal can be a principled and pragmatic response to the incursion of GenAI technologies in college writing courses.” What I like about this guide is that it acknowledges AI as shaping our writing environments, and then it provides resources to actively work against AI’s negative influence. Further resources for this approach are available in Refusing, Rejecting, and Rethinking Generative AI in Writing Studies and Higher Education, curated by Maggie Fernandes, Megan McIntyre, Kat Gray, and Cara Marta Messina.

Other courses might adopt an AI minimalism approach. Dan Cryer, who teaches at Johnson County Community College outside Kansas City, offers what he calls “critical minimal adoption” of AI. His approach asks, “What’s the least I can responsibly do with AI in my teaching?” He urges teachers to integrate AI “creatively and responsibly” and provides examples of teaching against AI as well as teaching with AI. We need to acknowledge AI and have conversations with students, but AI training is not our job. He argues—persuasively, I think—that businesses should handle any AI training, and writing teachers should instead focus on what we already do well: teach rhetorical judgment, critical reading, and writing.

Finally, some courses might embrace AI integration. These courses would actively encourage students to explore AI tools and applications. What’s great about this approach is that students can do original research on AI, share it with the class, and even faculty can learn from the work. My Writing Machines class at Pitt, co-developed with Matt Burton in our School of Computing and Information, asks students to explore co-creation with AI—and I’ve learned as much as them each semester I’ve taught it. This course counts for both our Public and Professional Writing major and our Digital Narrative and Interactive Design major at Pitt, which both emphasize hands-on experience with digital composing.

Also embracing AI, Tim Laquintano at Lafayette College offers an advanced professional writing class, Writing and Artificial Intelligence, which asked students to experiment with prompts and settings in LLMs, build a bot representing their approach to work in the class, and compete with AI for changing the reading level on policy documents. The work of that class is less focused on product and more focused on students’ critical assessment of their own and AI’s capabilities.

To document embracing AI in general Composition classes, Pamela Baker at University of Central Florida asks for a detailed GenAI Disclosure Statement, which links to their conversations so she can track their prompts and processes. Not all of her assignments allow for AI, but for any use of AI, students are required to submit this statement. She told me that this information from students is invaluable: “Having those inputs and outputs has allowed me to see how students are using it, which LLMs they're choosing, and trace their thinking around its use.” She’s seen students begin to use AI more critically because of this open approach—and even admit that using AI is less exciting once it’s not forbidden!

Whatever the approach, being AI aware means being deliberate about the integration of AI, and aligning that integration with our stated learning goals, so that we can help students make good choices.

In the classroom

You can work on your deliberate approach to AI integration by doing an AI-aware audit. First, look at the goals of your program, course, or assignment. Work with colleagues if you can—this kind of work is better done collaboratively! Break down your goals and outcomes into the smaller tasks used to achieve each, and ask: how does the presence of AI in our writing environments change the strategies and pathways we’ve used to get to this goal? Should we revise our goals? revise our strategies? How can we achieve these goals using pathways that acknowledge AI?

Here are some questions for conducting an AI-aware audit of your program, course, or assignment:

• What are the stated student learning outcomes?

• Why are these the goals, and should these goals be reaffirmed or revised?

• Where do students go after graduation, or this course, or this assignment?

• What do you and your faculty colleagues find important and valuable about this program, course, or assignment? What do students appreciate? A survey of students in a particular course or major can help to get these answers, if your program doesn’t already know them.

Then ask some questions related to AI:

• How does AI change the pathways to those student learning outcomes?

• How can assignments be revised in light of AI’s capabilities?

• What program-wide policies will encourage students to make responsible choices about AI use?

• What AI skills do students need to know?

• What additional assignments may be needed to get students to reach these goals?

• Are there additional classes that should be added to the curriculum to meet these needs?

MIT Sloan Teaching & Learning Technologies offers an excellent resource, 4 Steps to Designing an AI-Resilient Learning Experience, which echoes some of my advice and includes a toolkit of additional questions.

Share your examples of refusing, minimizing, or embracing AI in the comments!

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more ideas for AI-aware approaches to teaching writing.

Subscribe now

The ubiquity and pliability of the course makes it a catchall for a variety of skills and new, so-called literacies. Those of us who teach FYC know how important the course is to students’ writing development, but it’s a struggle to address everything. And now, here comes “critical AI literacy.” Can we fit that into the already overstuffed drawer that is FYC? Should we? What is critical AI literacy, anyway?

I think we should be teaching critical AI literacy in FYC or other writing-heavy gen-eds. Here, I’ll explain why, along with some ideas for how to do it.

Why teach about AI in writing classes?

On the one hand, some uses of AI undercut the goals of our writing classes. AI promises to summarize, write, research, rewrite, and respond—much more efficiently than human writers and readers. But efficiency and outsourcing are counter to the work we ask students to do as writers. On the other hand, students are already using AI, sometimes even if we ask them not to.

Plus: AI is rapidly being integrated into professional writing contexts. And large-scale public and private investments coupled with a permissive regulatory environment suggest that AI will not be extricated from these professional contexts that we’re helping prepare students to enter any time soon.

Of course, there are good reasons not to simply kowtow to the tech industry, power, money, the concept of college as vocational training, and so on. The Refusing GenAI in Writing Studies guide offers more justifications and strategies on that point (which I’ll say more on in a future post!).

Yet it’s clear that students will need to know something about AI. Teaching students critical AI literacy in at least some of our writing classes—a place they’re already using it—can help them to be better equipped to responsibly encounter, use, and resist AI. The first step to teaching critical AI literacy, of course, is figuring out what it is.

What is critical AI literacy?

“Literacy” tends to be a catchall term for any skill deemed important. We should all have textual, digital, financial, emotional literacy, and so on . . . ! I admit I’m skeptical of claims for everything being a literacy. But “critical AI literacy” has gotten traction, and it’s a good concept for considering how we go about teaching AI in our writing courses.

For me, critical AI literacy is the ability to understand, apply, and assess AI operations, uses, and outputs. This is how I think of it when breaking the concept down for faculty and students:

• Functional AI literacy: Can you use AI to accomplish tasks? Can you support your own learning with AI? Do you know how it works?

• Rhetorical AI literacy: Can you evaluate AI outputs? Can you tailor its uses and operations to your personal and community values and needs? Do you know when it might be appropriate and inappropriate to use AI?

• Ethical AI literacy: Do you understand the risks of AI, including environmental, societal—and even existential risks? Do you understand your own personal risks?

I’m building here on Stanford University’s AI literacy framework; I’ve added my own questions to Stanford’s organizing terms to help refine what I like to emphasize in classes.

Students should know how to use AI well—have some literacy in it, if you will. They should also be able to discern when to use it, when to avoid it, and how to evaluate it—that’s the critical part of “critical AI literacy.” Maha Bali has a great breakdown of the three parts of critical, AI, and literacy, noting that “critical” means both questioning and critiquing.

The MLA/CCCC Joint Task Force on Writing and AI also has a working paper on critical AI literacy that calls on educators to shape the culture of AI use by teaching it responsibly. Every agent in higher ed, from students and faculty to departments and administrators, should aim for greater literacy in AI. They provide a detailed list of what this literacy entails, which includes understanding: bias, data security, citation, limitations of AI, technology change over time, potential impact on learning, and determining appropriate uses of AI. And they offer specific advice for FYC and ESL programs, career readiness programs, academic integrity offices, and more. The document has been critiqued for accepting the inevitability of AI (see comments on the document itself), but I find it to be a thorough, thoughtful, and helpful guide for working toward greater AI literacy.

In the classroom

Ok, fine, some of our writing classes (not all!) should include attention to critical AI literacy. But what does that look like? Here are three activities I’ve used. They can be adapted for writing classes at any level, and you can run them even if you’re totally new to AI. These alone won’t fulfill a critical AI literacy student learning outcome (SLO), but they’re ways to begin.

• Can you evaluate AI outputs? A comparative exercise can help students see for themselves the difference between large language models (LLMs), and what rhetorical choices they make in summarizing text. My favorite way to do this: in class, enter the first paragraph of the Declaration of Independence into the site Chatbot Arena, and ask for a one-sentence summary. Chatbot Arena pits two random LLMs against each other in responding to a single prompt, and it allows users to rate the results. It also keeps track of ratings on a leaderboard. Students can compare what’s in the original paragraph and the summary sentences. Most LLMs change “men” to “people,” erase reference to “the Creator,” and they all offer different emphases on what’s in the original text. Chatbot Arena requires no login and it’s easy to use, plus its LLM leaderboard offers another perspective on how LLMs are evaluated. This summary exercise was designed by Tim Laquintano and is available as a free resource here.

• When is AI appropriate to use? In my experience, students have an inherent sense of when it’s inappropriate to use LLMs and often welcome open discussions about their choices. One way to start this conversation is to have students view the Gemini 2024 Olympics ad, “Dear Sydney,” alongside the Google Pixel 2025 Superbowl ad, “Dream Job.” Both ads feature dads using the Gemini LLM with their daughters, but the “Dear Sydney” ad has the dad ghostwrite a fan letter while the “Dream Job” ad features a dad prepping for an interview by talking about fatherhood. (Some teachers have assigned my blog post on the “Dear Sydney” ad and found it useful for context.) These ads are pretty simple to grasp, and discussing them helps get students thinking about how contexts for writing and AI differ, and how using AI might shape our personal relationships in various ways.

• What are some risks of relying on AI outputs? Students have heard that AI can be biased, but they might not have recognized examples of that bias. I’ve started discussions with students by screening Joy Buolamwini’s AI, Ain't I a Woman?, a powerful 3-minute performance about errors in AI facial recognition. First, I introduce Sojourner Truth’s speech, “Ain’t I a Woman?”—which many students are unfamiliar with. I then mention that AI facial recognition is a controversial technology, which Dr. Buolamwini helped to draw attention to with her personal experience and research. We watch the video, and students immediately get the problem with the way that AI classifies some faces better than others. To apply these concepts of recognition and representation, I’ve also asked various LLMs to “give me a picture of a university professor” or “give me a picture of a [gender studies / engineering / writing] university professor” and showed students the results to demonstrate how LLMs have different defaults. You can ask students to prompt different LLMs to generate images for “carpenter” or “nurse” and talk about their results. Ask: Who do you see represented in these images, and why? What datasets and concepts do you think the LLMs are drawing from, and how do you think their outputs are shaped by the parent company’s decisions? What are the implications of these representations—both in the United States, and globally? Here are some slides I’ve used for discussing bias in AI.

Reflection is key in any FYC activity, so I always try to help students process what they’ve learned in discussion or writing. AI gives us a great opportunity to try new things—so don’t be afraid to learn alongside your students.

Thanks for reading AI & How We Teach Writing! Subscribe and come back to hear more ideas for bringing critical AI literacy into the classroom.

Subscribe now

On Valentine’s Day in 2019, a little-known company called OpenAI published a blog post announcing GPT-2, an AI language model that could do: “rudimentary reading comprehension, machine translation, question answering, and summarization.” Back then, GPT-2 was merely the next step in one AI research company’s agenda—a leap, to be clear! But it seems no one intended to disrupt higher education, at least not at first. Yet less than three years later, OpenAI launched ChatGPT, which, among educators in 2025, needs no introduction.

What was remarkable about GPT-2 wasn’t that it could perform those reading, writing, and translation tasks, but that it could do them without specific training. Previous models were trained to answer specific kinds of questions or to comprehend narrow domains. GPT-2 was, instead, given 40GB of text from 8 million webpages on the internet and then “trained simply to predict the next word” in that text. The result was a general-purpose language model that could handle new texts and tasks, even ones its designers had never thought of. That’s why it’s called a GPT: generative pre-trained transformer. Using transformer technology, the model generates text based on training that precedes any specific task. Put another way, GPT-2 was the first language model that reads and writes pretty much anything you want.

What’s interesting, or perhaps harrowing depending on how you feel about it, is that this announcement of GPT-2 offers an alternative history to our current timeline. Just after OpenAI describes the model’s capabilities, they write:

Due to our concerns about malicious applications of the technology, we are not releasing the trained model. As an experiment in responsible disclosure, we are instead releasing a much smaller model for researchers to experiment with, as well as a technical paper.

OpenAI feared what GPT-2 could do in the wrong hands: generate fake news; impersonate others; automate spam or abuse through social media. They knew, in other words, that this powerful language model was potentially dangerous, and so they weren’t going to release it.

As a writing researcher interested in technology, I was excited by this choice. Look! The technologists see how powerful writing can be! Automated writing through AI language models could be used maliciously, but OpenAI also noted that this technology could improve dialogue assistants and speech recognition for accessibility or could provide translation between languages. I saw that potential, too.

It all seems kind of quaint now. Because eight months after OpenAI said they weren’t releasing the full model, they released the full model. They’d been under intense pressure to do so, with critics saying that they were overstating the risks, stifling innovation, and wasn’t their name OpenAI?? The release of GPT-2 didn’t make the splash that ChatGPT did two years later because you still needed to go through some extra steps to access the model. But it was a public launch of a powerful language model that carried significant risk—and promise.

OpenAI titled GPT-2’s launch post “Better language models and their implications.” Since this newsletter explores the implications of AI for teaching writing, I figured this was a good place to begin.

About this newsletter

I’m not an AI enthusiast. When I run a workshop on AI and teaching for writing faculty, I often say I wish I had a magic wand that would take us back to the time before ChatGPT. As a writing researcher, I find it thrilling to watch this writing technology roll out. What a time to be alive! AI large language models are the most rapidly adopted writing technologies in the history of the world! But as a writing teacher, I’m much less enthusiastic. These technologies promise efficiency and ease—which are not values I emphasize in a writing class. AI forces us to rethink and revise so much of our teaching. And I’m still tired from pandemic virtual teaching.

But here we are. We didn’t ask for this disruption, but we’re all in it together. One of the things I did as a Writing Program Administrator (WPA) during the pandemic was to call colleagues together to tackle our collective challenge. My hope is that this newsletter similarly helps us to navigate AI and how we teach writing. I’ve been living with AI and writing for a few years. I’m glad to share what I’ve learned, plus learn more from you in this space.

Twice a month, I’ll write about some contemporary aspect of AI that’s relevant to teaching college writing. Topics might include critical AI literacy, citing AI, how students are using AI, and how we might resist AI in our classes. I’ll always include some references and ideas for how you might bring the conversation into your classroom. I teach graduate and undergraduate classes at a large, state-supported, research-intensive university—University of Pittsburgh—and I started my teaching career in high schools. That’s my context, but of course, yours will be different. You are welcome to adapt anything I share here to your own teaching.

I’ve been writing on Substack for a year now, and I like the format for publishing timely pieces on AI. (You can check out my posts here.) In the middle of my career and deep into my second book, these posts remind me that I love to write. Writing is a mode of thinking—and isn’t that why we’re here? So, when Norton asked me about launching a newsletter on AI and writing to support teachers and their students, I couldn’t say yes fast enough! I’m excited to work with Norton because I get to connect with more teachers about AI. I hope that you feel welcome to share your experiences and ask questions in the comments. Please reach out to me directly if you have a specific question or an idea for a topic you feel should be covered here.

In the classroom

And here’s the part of the newsletter where I offer a few ideas for how to take this conversation about AI into the classroom.

In my upper-level writing classes focused on writing technologies, I’ve assigned the OpenAI blog post introducing GPT-2 in 2019, “Better Language Models and their Implications.” Students find it an interesting historical document for a technology they’ve all used. The post gives examples of how the model completes text-based prompts, comparisons to other contemporary models, and examples of how this model could be used as assistive technology or for malicious purposes.

Here are some questions I’ve asked in discussion:

• Was it responsible for OpenAI to release this smaller model and hold back the more powerful one?

• What if OpenAI had continued to keep the more powerful model private?

• Which implications of “better language models” does the post get right, and which ones does it miss?

Much of our current moment is foreshadowed in this post, but OpenAI misses some things, too. For instance, they mention nothing about education in their original post in Feb 2019, or the follow-up in Nov 2019 that announced the full model release.

If you’d like to extend the discussion into hands-on work with current language models, students can input versions of these questions into publicly accessible language models such as ChatGPT, Gemini, and Claude. I asked these three models, “In 2019, was it responsible for OpenAI to release the smaller model of GPT-2 and hold back the more powerful one?” The differences in their answers reveal variances that could spark conversation about the styles and values they express.

Students who want to learn more about GPT-2 can read its explanatory model card. Those who want to work on their technical skills can access GPT-2 on Hugging Face. An article from the Verge from Feb 14, 2019, “OpenAI’s new multitalented AI writes, translates, and slanders,” highlights the significance of GPT-2’s release and provides contemporary context.

Showing the history and contingency of these technologies helps students to see that they aren’t inevitable and that humans are behind their design and deployment. Knowing this about language models is part of critical AI literacy.