Have you ever asked an artificial intelligence tool to help you create a learning activity, only to receive something so generic that you could have written it yourself in half the time? Perhaps you requested a discussion prompt and received a question that encouraged little more than summarizing the textbook. Or maybe you asked AI to develop learning objectives and ended up with a collection of vague statements about students "understanding" or "appreciating" the material. The response looked polished, but it wasn't particularly useful.
If you've experienced this, you've encountered one of the most common challenges educators face when working with generative AI. These tools can produce impressive results, but the quality of those results often depends on how clearly we communicate what we need. A vague request leaves considerable room for interpretation, while a thoughtfully constructed prompt provides direction, context, and expectations. The difference can be substantial, particularly when developing instructional materials that must align with specific learning outcomes.
This is where prompt engineering becomes useful. Although the term may sound like something reserved for computer scientists, the underlying concept is surprisingly familiar to educators. Writing an effective AI prompt involves many of the same skills we use when developing clear assignment instructions, learning objectives, and assessment criteria. This article examines how educators and instructional designers can write better prompts, evaluate AI-generated responses, and use artificial intelligence as a practical assistant without surrendering the professional judgment that makes good teaching possible.
What Is Prompt Engineering?
Prompt engineering is the process of designing and refining instructions that guide an artificial intelligence system toward a desired response. In generative AI tools, a prompt may be a question, a request, a set of instructions, or a combination of information and examples. The purpose is to communicate what the user wants the system to accomplish and provide enough context to produce a relevant result.
For educators, prompt engineering is less about mastering technical commands and more about communicating instructional intent. When we ask AI to create a lesson plan, discussion question, rubric, or assessment, we are asking it to perform a task that depends on numerous instructional decisions. The tool needs information about the learners, learning objectives, subject matter, instructional environment, and expectations. Without that information, it must make assumptions, and those assumptions may not match the course we are designing.
Think about the difference between telling a student to "write a paper about leadership" and providing a carefully developed assignment with learning objectives, expectations, evaluation criteria, and a clear description of the intended audience. The second approach gives the student a much stronger foundation for producing appropriate work. Prompt engineering operates on a similar principle, although AI systems do not interpret instructions or understand learning in the same way people do. Clearer directions can improve the relevance of AI-generated material, but the results still require careful review.
Why Vague Prompts Produce Generic Results
One of the most common mistakes educators make when using AI is assuming the tool already understands the instructional situation. A request such as "Create a discussion question about political participation" may produce a perfectly reasonable question, but it tells the system almost nothing about the course. Is the audience composed of first-year undergraduate students or graduate students? Should the discussion emphasize foundational knowledge, critical analysis, or application? Are learners expected to connect their responses to current events or assigned readings?
Without those details, AI often generates a broadly applicable response. It may produce something like, "Why is political participation important in a democracy?" While that question could serve as a starting point, it does not necessarily encourage the level of analysis or interaction the instructor wants. Students might respond with general statements about voting and civic responsibility without engaging meaningfully with course concepts.
A more developed prompt might ask AI to create a discussion question for first-year American government students that requires them to compare two forms of political participation, connect their analysis to a course reading, and consider how political efficacy influences individual participation. The resulting question is more likely to support the intended learning outcome because the prompt communicates the instructional expectations. The improvement comes from providing meaningful context rather than simply adding more words.
This distinction is important because longer prompts are not automatically better prompts. A lengthy request filled with unnecessary details can create confusion just as easily as a vague one. Effective prompt engineering focuses on supplying the information that genuinely affects the task while communicating expectations clearly and directly.
Begin With the Learning Objective
Instructional designers are accustomed to beginning course development with learning objectives, and the same principle applies when using AI. Before requesting an activity, assessment, or instructional resource, it helps to identify what learners should be able to demonstrate. That decision gives the prompt a clear instructional purpose and provides a standard against which the generated response can be evaluated.
Imagine asking AI to develop an activity about accessibility in online courses. Without additional guidance, the tool might suggest a general discussion about why accessibility matters. That could be useful for introducing the topic, but it would not adequately support an objective requiring learners to evaluate instructional materials for accessibility barriers. The activity and the intended performance would be misaligned.
A stronger request would explain that learners must evaluate a sample course page against selected WCAG criteria and recommend specific improvements. AI could then suggest a realistic scenario, sample materials, or guiding questions that support that performance. The instructional designer would still need to verify the accessibility guidance and evaluate whether the activity appropriately measures the objective, but the prompt would establish a much stronger starting point.
This approach also helps prevent AI from generating attractive but instructionally unnecessary activities. A creative activity is not automatically an effective learning activity. By identifying the learning objective first, educators can evaluate suggestions based on whether they support the intended outcome rather than whether they simply sound interesting.
Give AI the Context It Needs
Context is one of the most important components of an effective prompt. Educators work within specific instructional environments, and those environments influence what constitutes an appropriate response. A strategy that works well in a face-to-face graduate seminar may not work as effectively in an asynchronous introductory course with a large enrollment.
Consider an instructor developing an activity for an eight-week online course. The instructor may have limited opportunities for synchronous interaction, students may be balancing employment and family responsibilities, and the activity must fit within an already demanding schedule. If AI does not know those constraints, it might recommend a live group presentation or an extensive collaborative project that would be difficult to implement.
Providing context helps the system generate suggestions that better match the learning environment. Useful information might include the course level, learner characteristics, delivery format, available technology, time limitations, prerequisite knowledge, and instructional goals. These details help establish what is realistic and appropriate for the situation.
However, educators should distinguish between providing instructional context and sharing sensitive information. AI does not need access to identifiable student records, confidential evaluations, or protected institutional information to generate a useful learning activity. General descriptions of learner needs and instructional conditions are often sufficient. Responsible prompting includes knowing what information to provide and what information should remain private.
Be Specific About the Task
An effective prompt should clearly communicate what the AI system is expected to produce. Educators sometimes combine several unrelated requests into one instruction and then become frustrated when the response addresses some requirements while overlooking others. Breaking a complicated task into manageable stages can produce results that are easier to evaluate and revise.
For example, developing a complete course module involves several distinct decisions. The learning objectives must be appropriate, assessments must measure those objectives, activities must prepare learners for the assessments, and instructional materials must support the entire process. Asking AI to generate everything simultaneously may produce a polished-looking module, but it can also introduce alignment problems that are difficult to notice.
A more effective approach is to begin with one task and build upon the results. An instructional designer might first ask AI to review the clarity and measurability of existing learning objectives. After revising those objectives, the designer could request several assessment approaches that would provide evidence of the intended performance. Only after selecting an appropriate assessment would the designer move toward developing learning activities and supporting materials.
This staged approach resembles the instructional design process itself. Each decision informs the next, and the designer has opportunities to review the work before additional material is generated. AI becomes part of a deliberate development process rather than a system producing an entire course from a single request.
Tell AI What a Good Response Looks Like
Educators understand the importance of communicating expectations to learners. A well-developed rubric explains what successful performance looks like, while clear assignment instructions reduce uncertainty about the task. Similar principles can improve AI prompts because they establish criteria for evaluating the generated response.
Suppose an instructor wants AI to create multiple-choice questions for a course. Simply requesting ten questions may produce items that primarily assess factual recall, contain implausible distractors, or use unnecessarily complicated wording. If the instructor wants questions that assess application, the prompt should communicate that expectation directly.
The instructor might specify that each question should present a realistic scenario, require learners to apply a course concept, include four plausible answer choices, and provide an explanation of the correct answer. The prompt could also request that incorrect choices reflect common misconceptions rather than obviously unreasonable alternatives. These instructions give the system a clearer description of the desired output.
The same approach applies to rubrics, discussion prompts, case studies, and instructional explanations. Educators can identify the qualities that matter most and communicate those expectations before requesting a draft. This does not guarantee that AI will satisfy every criterion, but it makes the output easier to evaluate and refine.
Examples Can Be More Useful Than Explanations
Sometimes the easiest way to communicate an expectation is to provide an example. Educators routinely use models and exemplars to demonstrate successful performance, and AI systems can also benefit from seeing the kind of output a user wants. This technique is commonly called few-shot prompting when the prompt includes a small number of examples.
Imagine asking AI to write module introductions that match an established course style. You could describe the desired tone as professional, conversational, welcoming, and concise. Those instructions might produce acceptable results, but the system may still interpret the tone differently than intended. Providing one or two examples of existing introductions gives it a more concrete reference.
Examples are particularly useful when consistency matters. Instructional designers may want announcements to follow a familiar structure, learning objectives to use a specific format, or feedback comments to maintain a particular tone. Showing AI an appropriate example can help establish those expectations more effectively than relying on descriptive language alone.
However, examples should be selected carefully. If the example contains weak instructional practices, AI may reproduce those weaknesses along with the desired formatting. Educators should review exemplars before using them and remain attentive to whether the generated response preserves instructional quality rather than merely copying surface characteristics.
Ask for Alternatives Instead of Accepting the First Answer
One of the most productive ways to use generative AI is to request several possible approaches rather than expecting a single perfect response. Instructional design problems rarely have only one solution, and comparing alternatives encourages more thoughtful decision-making. AI can help generate possibilities quickly, allowing educators to consider approaches they might not have identified independently.
For example, an instructor might ask for three different ways to assess students' ability to evaluate political information. One suggestion might involve comparing news coverage, another might use a case study, and a third might ask learners to analyze claims made in a public communication. Each option could support the learning objective, but the instructional requirements and workload would differ.
The instructor can then evaluate the options according to course goals, available time, accessibility, and learner needs. Perhaps the first idea is too complicated for an introductory course, while the second offers a manageable opportunity for meaningful analysis. The third might become useful later in the semester when learners have developed additional skills.
This is where AI can function as a thinking partner rather than a content generator. The educator remains actively involved in comparing, questioning, and refining possibilities. The goal is not finding an answer that eliminates professional decision-making. It is generating options that make those decisions more informed.
Treat Prompting as an Iterative Process
Even a thoughtfully written prompt may not produce exactly what an educator needs on the first attempt. AI might misunderstand the intended audience, provide examples that are too advanced, overlook a constraint, or generate an activity that does not adequately support the learning objective. These problems do not necessarily mean the tool is useless. They indicate that the response needs evaluation and the instructions may need refinement.
Rather than starting over every time, educators can identify what needs improvement and provide targeted follow-up instructions. If a discussion prompt encourages too much summarization, ask AI to revise it so learners must compare alternatives and justify their reasoning. If a rubric contains overlapping criteria, ask for clearer distinctions between the performance dimensions. If an explanation is too technical for the intended audience, request a revision using more accessible language while preserving accuracy.
This process is similar to revising instructional materials after reviewing feedback. The initial response becomes a draft that can be examined, questioned, and improved. Over time, educators may also develop reusable prompt structures that consistently produce useful starting points for recurring tasks.
It is important, however, to know when continued revision is no longer productive. Sometimes an AI-generated response is fundamentally unsuitable, and rewriting the material directly may be faster than repeatedly prompting the system. Effective AI use includes knowing when the technology is helping and when it is creating additional work.
Use AI to Review, Not Just Generate
Much of the conversation about prompt engineering focuses on creating new material, but AI can also assist with reviewing existing instructional content. This can be particularly useful for instructional designers who spend considerable time evaluating course materials, identifying inconsistencies, and helping faculty improve instructional clarity.
For example, an instructional designer might provide a de-identified assignment description and ask AI to identify areas where students could reasonably misunderstand the instructions. The tool could suggest questions learners might ask, highlight ambiguous terminology, or identify places where expectations are not clearly communicated. The designer could then determine which suggestions warrant revision.
AI can also help examine alignment. Given a learning objective and an assessment description, the system may identify potential mismatches between the expected cognitive performance and the task students must complete. It might notice, for example, that an objective requires evaluation while the assessment primarily measures recall. That observation could provide a useful starting point for a deeper instructional review.
These uses are most effective when AI is treated as a second reader rather than a final reviewer. The system may identify issues worth considering, but it can also overlook problems or suggest changes that weaken the instruction. The educator remains responsible for evaluating the recommendations and determining whether they improve the learning experience.
Prompt Engineering Cannot Replace Subject Matter Knowledge
One of the risks of generative AI is that its responses often sound authoritative even when they contain errors. A well-written prompt can improve relevance and organization, but it cannot guarantee factual accuracy. AI systems may generate incorrect explanations, invent references, misinterpret policies, or present unsupported claims with considerable confidence.
For educators, this means prompt engineering must always be paired with verification. A generated lesson about a historical event should be checked against reliable historical sources. A proposed science activity should be reviewed for technical accuracy and safety. An accessibility recommendation should be verified against current standards and appropriate guidance. The quality of the writing should never be mistaken for evidence that the content is correct.
Subject matter expertise remains essential because educators need to identify what the AI system gets wrong. Instructional designers working with unfamiliar disciplines should collaborate with faculty or other qualified experts rather than assuming that a polished response is accurate. AI may accelerate drafting, but it does not eliminate the need for disciplinary review.
This is especially important when AI generates citations. References that look legitimate may contain incorrect publication details or refer to works that do not exist. Educators should verify sources independently before including them in course materials. A convincing citation is not necessarily a real citation.
Protect Privacy and Academic Integrity
Responsible prompt engineering also requires attention to privacy, institutional policies, and academic integrity. Educators may work with student assignments, grades, advising information, research materials, and confidential institutional documents. Not all of that information belongs in a generative AI system, particularly when the tool has not been approved for processing sensitive data.
Before sharing instructional materials with AI, educators should understand the institution's policies and the platform's data-handling practices. Personally identifiable student information and protected educational records should not be entered into unapproved tools. When possible, instructional tasks should use generalized scenarios, fictional examples, or appropriately de-identified materials that do not reveal confidential information.
Academic integrity also deserves consideration. Educators should be transparent about how AI is used in developing instructional materials and should establish clear expectations for student use when appropriate. AI can support learning, but it can also make it easier to bypass the thinking that assignments are intended to develop.
The goal is not avoiding AI simply because risks exist. It is making informed decisions about when and how the technology can be used responsibly. Good prompting begins with instructional intent, but responsible use also requires attention to the people and information affected by the process.
Better Prompts Begin With Better Questions
Perhaps the most interesting aspect of prompt engineering is how closely it resembles effective instructional design. Both require clarity about the desired outcome, an understanding of the audience, appropriate context, meaningful expectations, and a process for evaluating results. The same habits that help instructional designers develop strong learning experiences can also help them communicate effectively with AI systems.
Consider how often instructional designers ask faculty to clarify what students should actually be able to do. We encourage instructors to move beyond vague goals, identify observable performance, and develop assessments that align with those expectations. Prompt engineering asks us to apply similar discipline to our own requests. We must decide what we want, why we want it, and how we will determine whether the result is useful.
This also explains why experienced educators may be particularly well positioned to use AI effectively. They already understand how to structure instructions, provide examples, establish criteria, and evaluate performance. Those skills transfer naturally to writing better prompts, even though AI systems and human learners operate very differently.
Ultimately, the most useful prompts begin with thoughtful questions. What problem am I trying to solve? What information does the system need? What constraints matter? How will I evaluate the response? Answering those questions before writing the prompt often saves more time than any clever wording technique.
Final Thoughts
Prompt engineering for educators is not about discovering secret phrases that make artificial intelligence produce perfect answers. It is about communicating instructional goals clearly, providing appropriate context, defining expectations, and evaluating results with professional judgment. Better prompts can produce more relevant drafts, stronger brainstorming options, and more useful suggestions, but they do not eliminate the need for review.
For instructional designers, the greatest opportunity lies in using AI to support the work we already know how to do. We can ask it to suggest alternative activities, identify potential problems in assignment instructions, organize ideas, or help develop initial drafts. We can refine those outputs through follow-up questions and use them as starting points for better instructional decisions. Throughout the process, learning objectives, learner needs, accessibility, accuracy, and ethical considerations must remain central.
The most effective educators will not necessarily be those who write the longest or most complicated prompts. They will be the ones who understand their instructional goals well enough to ask meaningful questions and evaluate the answers they receive. AI can generate possibilities, but educators must determine which possibilities genuinely support learning.
Better prompts can produce better responses. Better instructional judgment is what turns those responses into better learning experiences.
References
Mollick, E. (2024). Co-intelligence: Living and working with AI. Portfolio.
OpenAI. (n.d.). Prompt engineering. OpenAI API documentation. https://platform.openai.com/docs/guides/prompt-engineering
UNESCO. (2023). Guidance for generative AI in education and research. https://unesdoc.unesco.org/ark:/48223/pf0000386693
U.S. Department of Education, Office of Educational Technology. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations. https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf
World Wide Web Consortium. (2023). Web Content Accessibility Guidelines (WCAG) 2.2. https://www.w3.org/TR/WCAG22/