Prompting for Educators

Jason Blomquist; Liza Long; Joel Gladd

5 Prompting for Educators

Where AI Fits in Faculty Work

Faculty do not need to begin with a full course redesign. A more manageable starting point is a teaching task you already understand well enough to judge. You know what a useful discussion question looks like in your field. You know when an assignment sheet is confusing, whether a quiz question matches the course level, and which parts of a rubric reflect what you actually value.

Research on faculty adoption supports this emphasis on judgment and pedagogy. In one mixed-methods higher-education study, educators identified a need for training in pedagogical strategies, evaluation of AI outputs, ethics, and adaptive expertise (Soleimani et al., 2025). A randomized trial that enrolled 259 secondary science teachers also found lower lesson-and-resource preparation time when teachers used ChatGPT with a practical guide: 56.2 minutes per week compared with 81.5 minutes in the comparison group (Roy et al., 2024). The teachers used AI for one or two bounded activities and had five weeks to become familiar with the approach before researchers measured planning time. The study involved secondary science teaching in England, so it does not establish a general higher-education workload effect. It does show why a narrow, recognizable task is a sensible place to begin.

Categories of Faculty Use Cases

Faculty work with AI tends to fall into a few broad categories. Each has different strengths and considerations.

Instructional Planning and Design

Instructional planning gives faculty several bounded tasks where an AI-generated draft can be compared with existing course goals and materials.

AI-Assisted Instructional Planning Tasks
Task	What AI Can Do	What You Still Do
Lesson planning	Generate outlines, suggest activities, propose discussion questions	Evaluate fit for your students, adjust timing, align with your objectives
Assignment creation	Draft instructions, suggest rubric criteria, generate example prompts	Ensure alignment with learning goals, calibrate difficulty, check for ambiguity
Quiz and exam questions	Generate questions across formats (multiple choice, short answer, essay prompts)	Verify accuracy, check for bias, ensure appropriate difficulty, review answer keys
Rubrics	Create criteria and performance level descriptions tied to your goals	Refine language, adjust weights, ensure criteria match what you actually value

The following prompt brings the R-T-C-I framework from Getting Started with AI Chatbots and Prompting into lesson planning. The Role is optional. The Task identifies the work, while the Context and Instructions specify the subject, level, class length, required elements, and limits.

You are an expert instructor in [subject]. Create a lesson plan on [topic] for a [level] course. Include: learning objectives aligned with [framework or standards], two active learning activities, and one brief assessment to check understanding. The class meets for 50 minutes.

A prompt without that context leaves the AI to guess at the audience and course conditions. The added context gives you a draft that is easier to compare with the class you actually teach.

Content Creation and Transformation

AI can help transform existing content into new formats or adapt materials for different audiences.

For many faculty tasks, a stronger prompt begins with the source material. If you want help revising a syllabus statement, upload the syllabus. If you want a clearer assignment sheet, provide the assignment sheet and rubric. If you want a plain-language policy explanation, provide the policy. Source material supplies local information the model may not otherwise have and gives you something concrete to check the response against.

Content Transformation Prompt Starters
Task	Example Prompt Starter
Summarizing	“Summarize this article in 200 words for students who have not read it…”
Simplifying	“Rewrite this explanation at a [grade level] reading level, avoiding jargon…”
Translating tone	“Rewrite this feedback in a more supportive, student-friendly tone while preserving the core message…”
Creating variations	“Generate three alternative ways to explain [concept], using different analogies…”
Formatting	“Convert this list of learning objectives into a table with columns for objective, assessment method, and class activity…”

AI-Assisted Feedback

Student feedback takes time and requires judgment about the assignment, the writer, and the course. AI-assisted feedback therefore needs a more careful process than asking a chatbot to comment on a paper.

Anna Mills presents AI-assisted feedback as one stage in a human-centered writing process. The PAIRR (Peer & AI Review + Reflection) approach begins with readings and discussion about AI risks, bias, language equity, environmental impacts, and hallucinations. Students then give peer feedback, request AI feedback on the same draft with instructor guidance about privacy, compare and evaluate both forms of feedback, and revise. The student’s goals, audience, writing voice, and judgment guide the process. Mills’s updated prompts show how much design this requires. The prompts include the assignment instructions and rubric, direct the system to ground comments in details from the student’s draft, limit feedback to a few priorities, acknowledge uncertainty, and avoid supplying language that a student could paste into the paper. Students decide which suggestions to accept, question, adapt, or reject. The 2025 PAIRR study by Sperber and colleagues examined this approach in ten courses with 654 students. The researchers found that AI and peer responses often reinforced or complemented each other. Students also practiced AI literacy by evaluating the feedback’s strengths and limits before revising. The evidence is specific to formative feedback embedded in peer review, reflection, and instructor guidance. The approach shows how AI might be used both critically and ethically to support feedback within a course.

Automated grading raises a different set of risks. Marc Watkins (2025) contrasts this kind of formative feedback with automated grading. He presents the approach as a process in which AI supplements peer feedback while students and educators retain agency and judgment. Faculty considering AI-assisted feedback should provide students with privacy guidance, define the tool’s role, and retain responsibility for evaluative judgments and final grades. Later chapters in this training address these ethical considerations more fully.

Sarah and the J-Curve

Sarah is an illustrative composite faculty member. Her ENG 101 Project from AI Tools and Platforms gave her a starting point, so she decided to push further with two tasks. First, she took her dense, two-page assignment sheet for the research paper and asked the AI to transform it into a student-friendly version. Her original instructions were written for herself as much as for students, full of conditional language and references to policies. The AI produced a cleaner version that preserved the requirements while making the expectations more readable. She still repaired a few places where the draft oversimplified the assignment, but the task was easy for her to evaluate.

Second, she tried generating a rubric for peer review. This attempt took longer. She kept adjusting the criteria, the performance-level descriptions, and the language until the output matched what she actually valued in student peer review. By the time the draft fit her course, writing the rubric herself might have been faster.

Sarah kept the instructions as a reusable template. When she later needed a rubric for a different assignment, she swapped in the new learning outcomes and task description. The new draft required fewer revisions because the template already reflected her priorities. Her experience illustrates the J-Curve without guaranteeing it: setup and experimentation can create an initial dip, while a well-chosen recurring task may become easier after the workflow is calibrated.

Ethan Mollick (2024) describes how reusable templates can carry an instructor’s priorities from one task to another. Sarah’s template included the structure she wanted, the kinds of criteria she valued, and instructions about tone and performance levels. She still reviewed each new rubric because a reusable template is a starting point, not a course decision.

Course Communication and Student Support

Some AI uses sit close to teaching without becoming assessment. These are tasks where the instructor already knows the message and uses AI to improve clarity, organization, or tone. Keep them low-risk: do not include confidential student details, do not outsource evaluative judgment, and do not use AI to manufacture personal advocacy.

Course announcements. Ask AI to review an announcement so students can quickly see what is due, where to submit, and what to do if they are stuck.

Review this course announcement to make sure it’s clear and easy for students to act on. Keep the tone friendly but direct. Preserve the deadline, submission location, and extension policy exactly as written. Add a short bullet list of the next steps students should take.

Assignment clarification. Give AI a confusing assignment prompt and ask it to identify where students may need more guidance. Then decide which suggestions fit your course.

Policy explanations. Use AI to turn syllabus or course-policy language into a plain-language FAQ. Check the result carefully against the original policy before sharing it.

Accessibility revisions. Ask AI to simplify dense instructions, generate alt-text drafts for instructor-created images, or suggest headings for a long handout. You still check the final version for accuracy and fit.

Student Engagement Tools

AI can help create resources that support student learning outside of class.

Practice problems. Generate additional examples for students who want more practice.

Study guides. Create review materials organized around key concepts and potential exam questions.

Tutoring chatbots. Design a prompt that turns the AI into a subject-specific tutor, guiding students through problems without giving away answers immediately.

Act as a patient and encouraging [subject] tutor. When a student asks a question, do not give the full answer immediately. Instead, ask clarifying questions to understand what they already know, then guide them toward the answer with hints and scaffolding. If they struggle, provide increasingly specific help. Always end with a question to check their understanding.

A 2025 randomized controlled trial at Harvard compared a custom AI tutor with an active-learning class in introductory physics (Kestin et al., 2025). Students using the tutor learned more on the measured lessons while spending less median time on task. The tutor was carefully engineered: it used expert-written prompts, structured sequencing, detailed solutions supplied by the designers, and established learning principles. The researchers also caution that their result may not transfer to settings requiring complex synthesis or higher-order critical thinking. The generic tutoring prompt above is a starting point for experimentation, not the system tested in that study.

In an interview published by Lance Eaton, instructional designer Jason Bock reported that students initially liked a constrained AI tutor but soon compared it unfavorably with open-ended tools such as ChatGPT (Eaton, 2025). This is one project account, not evidence of a general student preference. It does suggest that students will notice a tutor’s limits and may need a clear explanation of why those limits support the learning task.

Adapting Prompts to Your Discipline

Generic prompts often produce generic results. Disciplinary context gives the AI more information about the course, the students, and the standards you will use to evaluate the draft.

Consider what makes your teaching context distinct. What terminology do students need to learn? What common misconceptions do they bring? What types of problems or texts are central to your field? What standards or competencies guide your curriculum?

A prompt becomes discipline-specific when it reflects how the field formulates problems, uses evidence, and judges a satisfactory result. Stolpe, Larsson, and Johansson Falck’s (2026) theoretical framework for discipline-specific AI literacy describes four connected actions that faculty can use when planning an AI interaction:

Recognize the need. Identify a learning or teaching problem that AI might meaningfully support.
Articulate the problem. Frame the request using the field’s concepts, evidence, language, and constraints.
Contribute to learning. Use AI to test an explanation, explore alternatives, or refine a judgment while preserving the reasoning the learner needs to practice.
Analyze the consequences. Examine what the interaction strengthens, weakens, skips, or changes in the discipline’s ways of learning and producing knowledge.

These actions shift attention from the wording of a prompt to the disciplinary practice surrounding it. A detailed prompt can still produce weak teaching material if the task itself asks AI to bypass the judgment students are supposed to develop.

The open-access, peer-reviewed Writing and Rhetoric Studies in the Loop: A GenAI Prompt Library shows what this looks like in one field. Edited by Anuj Gupta and Brian Gogan, the collection presents 16 prompts built from writing and rhetoric concepts and tested across multiple generative AI platforms. Each entry includes the disciplinary idea behind the prompt, the complete prompt, sample outputs from three platforms, and contributor notes that invite readers to examine the design choices and results. Faculty in other disciplines can use that structure as a model: identify the field knowledge the AI needs, encode it in the prompt, compare outputs, and decide what the tool still misses.

The AI Tools for Teachers handbook takes a workflow-library approach in a New South Wales K-12 context. It organizes 108 tasks across nine sections covering the practical work of teaching and school administration. Each task pairs an example prompt with practical steps and tool options, followed by reminders about privacy, source verification, and teacher judgment. Its policies and examples require adaptation for higher education. The reusable pattern is to begin with a recurring task, provide the relevant local source, specify the teaching context, and identify what the educator must still verify.

Embed these details in your prompts. A request like “create a case study” will get you something generic. A request like “create a case study involving a small business facing a cash flow problem, appropriate for an introductory accounting course where students have learned about financial statements but not yet covered ratio analysis” produces something you can actually use.

Source grounding is part of disciplinary adaptation. A prompt becomes much stronger when it includes the actual materials you want the AI to use: a reading, lab handout, dataset, clinical scenario, assignment sheet, rubric, transcript, institutional policy, or sample student-facing explanation.

Use only the attached source materials when answering. Quote or cite the specific passages you rely on. If the sources do not support an answer, say that clearly before suggesting a revision.

Disciplinary context also shapes what counts as a useful result. A case study that works in accounting may fail in nursing, history, welding, or philosophy because each field uses different evidence, conventions, safety boundaries, and forms of judgment. AI performance can also vary across tasks that appear similar. In a field experiment with 758 knowledge workers, Dell’Acqua and colleagues found gains on tasks within the tested system’s capabilities and lower correctness on a selected task outside them (Dell’Acqua et al., 2026). Faculty therefore have to test the task in their own context.

Your expertise belongs throughout the process. You select the task, provide the course context, inspect the draft, and decide whether revision is worthwhile. AI may generate a rough draft, offer alternatives, or handle formatting, but you determine whether the result fits the discipline and the students in front of you.

Try This: Discipline-Specific Adaptation

Take one of the sample prompts from this chapter. Adapt it for your specific teaching context by adding

Your course level and student background
Specific topics or concepts you’re covering
Any constraints (time, format, materials available)
The disciplinary practice students should use, such as interpreting evidence, diagnosing a problem, constructing an argument, applying a safety standard, or explaining uncertainty
The sources, conventions, terminology, or professional standards the result should follow
The part of the reasoning that faculty or students must still perform and defend

Run both the generic version and your adapted version. Compare the outputs. Which is more immediately usable? What changed when you added disciplinary context? Did the AI support the intended disciplinary practice, or did it perform the thinking students need to practice? Which parts would still need revision before you could use the material with students?

Setting Realistic Expectations

Sarah started with an assignment sheet and a rubric. She knew both tasks well enough to compare the AI drafts with her own past work. She also learned that a quick result on one task did not predict a quick result on the next.

Ethan Mollick (2024) calls the basic approach “good enough prompting.” Begin with enough specificity to produce a draft you can evaluate, then refine it. Alberto Romero (2026) argues that recognizing when an output is good enough is itself a learned ability. Another round of revision may help, or it may consume attention without improving the material enough to justify the effort.

Pick one or two recurring tasks and focus on those. Expect iteration and review. If a set of instructions works, you may reuse it as a starting point while changing the course content, audience, and constraints. Pay attention to the tasks AI handles poorly. A workflow that repeatedly requires more repair than creating the material yourself is a poor fit for AI right now.

Measured gains on one task do not automatically reduce total workload. The EEF trial measured a defined slice of lesson preparation. Dell’Acqua and colleagues found that AI improved performance on some knowledge-work tasks and reduced correctness on another. The evidence supports a narrower expectation. A bounded task may become faster or produce a stronger draft, while total faculty workload remains a separate question.

Processes Over Prompts

Prompts are one part of working with AI. The surrounding process determines whether a generated draft becomes usable teaching material.

Leon Furze (2025) makes this argument directly. Modern tools can accept documents, links, images, spreadsheets, and other forms of context, so faculty work increasingly involves choosing sources and managing a sequence of decisions. Choose a bounded task. Ground the request in course context or source material. Inspect the result against your standards. Revise it or stop.

That process should include a source decision. Before asking for a study guide, policy FAQ, assignment revision, or tutoring prompt, decide whether the AI should answer from general knowledge, current web sources, or materials you provide. When accuracy and local fit matter, ground the response in a source and then check whether the answer actually follows from it.

Practical fluency begins with a workable prompt and a process that keeps faculty expertise at the center. You decide what the tool should attempt, what information it can use, whether the result fits the course, and when the task is finished.

Looking Ahead

This chapter covered how faculty can apply prompting skills to real teaching, course communication, and student-support tasks. The next chapter, Agentic AI, introduces systems that can take several steps toward a goal, use tools, and pause for human review. You will consider where these systems fit faculty workflows and how prompting changes when the instructor must supervise a longer process.

References

Brynjolfsson, E., Rock, D., & Syverson, C. (2021). The Productivity J-Curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333-372. Source link

Dell’Acqua, F., McFowland, E., III, Mollick, E., Lifshitz, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science, 37(2), 403-423. DOI record

Roy, P., Poet, H., Staunton, R., Aston, K., & Thomas, D. (2024, December). ChatGPT in lesson preparation: A Teacher Choices trial. Education Endowment Foundation. Source link

Eaton, L. (2025, November 7). The concern is for average work. AI + Education = Simplified. Source link

Furze, L. (2025, November 3). Processes are more important than prompts. Source link

Gupta, A., & Gogan, B. (Eds.). (2026). Writing and rhetoric studies in the loop: A GenAI prompt library. The WAC Clearinghouse. Source link

AI tools for teachers. (n.d.). Retrieved July 20, 2026, from Source link

Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: An RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15, 17458. Source link

Mills, A. (2026, January 27). AI feedback customized for student writers: The updated PAIRR prompts. Anna Mills’ Substack. Source link

Mollick, E., & Mollick, L. (2024). Stop writing all your AI prompts from scratch. Harvard Business Publishing Education. Source link

Mollick, E. (2024, November 24). Getting started with AI: Good enough prompting. One Useful Thing. Source link

Romero, A. (2026, February 21). The most important skill in AI right now: How to know when to stop. The Algorithmic Bridge. Source link

Sperber, L., MacArthur, M., Minnillo, S., Stillman, N., & Whithaus, C. (2025). Peer and AI review + reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition, 76, 102921. DOI record

Soleimani, S., Farrokhnia, M., van Dijk, A., & Noroozi, O. (2025). Educators’ perceptions of generative AI: Investigating attitudes, barriers and learning needs in higher education. Innovations in Education and Teaching International, 62(5), 1598-1613. Source link

Stolpe, K., Larsson, A., & Johansson Falck, M. (2026). Discipline-Specific AI literacy (DiSAIL): A theoretical framework for situated engagement with generative AI in education. International Journal of Technology and Design Education, 36, 1917–1932. DOI record

Watkins, M. (2025, October 10). The dangers of using AI to grade. Rhetorica. Source link

License

Icon for the Creative Commons Attribution-NonCommercial 4.0 International License

A Guide to Teaching and Learning with Artificial Intelligence Copyright © by Jason Blomquist; Liza Long; and Joel Gladd is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, except where otherwise noted.