Generative AI in Instructional Design: What the Professionals Actually Do

When 70 instructional designers were asked how much they value generative AI for their work, they were enthusiastic. When the same people were asked how often they actually use it, the numbers dropped. That gap, between the promise and the practice, is the most useful thing in a new study by Tian Luo, Pauline Muljana, Xinyue Ren, and Dara Young, and I think teachers should pay close attention to it.

Instructional designers are worth listening to here. They’ve been building learning experiences for a living, with real deadlines and real stakeholders, since before ChatGPT existed. When a whole profession of design experts adopts a tool cautiously and names exactly what it’s good and bad for, that’s a signal. This post is about what that signal tells the rest of us who teach.

The Gap Between Enthusiasm and Everyday Use

Luo and colleagues ran a mixed-methods study, surveying 70 working designers and interviewing 13 of them. ChatGPT dominated the results: sixty percent called themselves extremely familiar with it and just over half used it daily, while the purpose-built education tools like Khanmigo and Education Copilot barely registered.

The headline pattern is a split I keep seeing everywhere in education. Designers were optimistic about what generative AI could do across nearly every part of their work. Their reported use told a more restrained story: low or infrequent in most areas of the actual job.

Where they did use it, they used it early. They generated first-draft objectives, drafted content, built out design documents, and bounced around ideas when stuck. In the later phases, implementation and evaluation, adoption fell off. The enthusiasm was broad. The real use was narrow and front-loaded.

Where it did earn its keep, the wins were concrete. Designers leaned on it as a kind of design partner for the messy, unstructured problems the job throws up, and to smooth collaboration with the subject matter experts they build courses around.

One designer used it to turn a nervous, non-native-English expert’s script into thirty usable videos. Another translated a dense technical email into something faculty could actually read. These are real gains, and they cluster around drafting, adapting, and communicating, not deciding.

I find that reassuring, not disappointing. These are professionals matching a tool to the tasks it genuinely helps with and holding it back from the ones it doesn’t. That’s discernment, and it’s exactly what we want students and teachers to develop too.

Exploring instructional designers' utilization and perspectives on generative AI tools- A mixed methods study

A Better Project Starter Than a Project Finisher

One designer in the study gave the whole thing its best line, calling AI a better project starter than a project finisher. Luo et al. found this sentiment nearly everywhere. Almost every designer flagged GenAI output as potentially biased, inaccurate, or generic, and many caught hallucinated citations and invented facts in what the tools handed back.

So the pros treat AI output as raw material, not a finished product. They generate with it, then they cut, correct, and rewrite. The value shows up at the start of the work, and the human judgment shows up at the end.

That maps almost perfectly onto an argument I made recently about frictionless AI: the real risk is letting AI carry the parts of the work that were supposed to build your thinking. [interlink: frictionless AI post]

There’s a clear classroom lesson in that. If we teach students to treat AI as a first-draft engine they are then responsible for finishing, we’re teaching the exact stance these professionals arrived at on their own.

Prompt Engineering Is a Literacy, Not a Trick

Nine of the designers named prompt engineering as a skill you have to build, with time and training, to get anything usable. It isn’t a hack or a magic phrase. It’s a competence you develop.

This one is worth pausing on, because it reframes what “using AI” even means. The people getting real value from these tools weren’t the ones who typed a question and accepted the answer.

They were the ones who learned to interrogate the tool, refine the request, and evaluate what came back. That’s a literacy, and it’s the same literacy I keep arguing we need to build with students on purpose, not just assume they’ll absorb it. [interlink: Med’s critical AI literacy work]

The uncomfortable implication is that access is not competence. Handing a teacher or a student a ChatGPT login does nothing on its own. The skill is in the using, and the skill has to be taught.

Concretely, that means building prompt craft into assignments the way we already build in citation or revision: show students a weak prompt beside a strong one, have them compare the outputs, and grade the quality of the questioning, not just the answer that comes back.

What Generative AI in Instructional Design Teaches Every Educator

The concerns these designers raised read like a preview of the conversations happening in every school. They worried about data privacy, knowing their prompts were traveling to servers they didn’t control.

Authorship and plagiarism came up just as often, with designers unsure when to disclose that AI had helped. And they pointed to something teachers feel acutely: a near-total absence of institutional policy to guide any of it.

That policy vacuum is the part I’d underline. One designer described the frustration of subject matter experts using AI to write the very lectures they were paid to produce, with no rules about whether that’s acceptable. Sound familiar? It’s the same question teachers are wrestling with about student work, just pointed in a different direction.

The answer is a clear, local policy that names what’s allowed and what has to be disclosed, not a blanket ban and not a blanket permission, which is exactly what I argued for when writing about course-level AI use. [interlink: Med’s course-level AI policy article]

The privacy worry translates just as directly. These designers stopped short of pasting sensitive company data into tools they didn’t control, and that instinct is a teachable one. Students and teachers should know that what goes into a public chatbot can leave their hands, which is reason enough to pause before feeding it a draft full of personal information, unpublished work, or anything a school is obligated to protect.

The Judgment Is the Job

One finding lands harder than the rest. When these designers hit the limits of the tools, the bias, the errors, the flat generic voice, they didn’t step back from their expertise. They leaned into it harder. Luo et al. report that IDs overcame those limits by exercising their analytical and critical thinking even more, and that they did not believe the tools would replace their skills.

That’s the whole argument in one finding. Generative AI didn’t make the judgment of these professionals less necessary. It made it more necessary. The tool got faster at producing a draft, which raised the value of the human who could tell a good draft from a bad one.

I’ve said for years that AI should amplify human thinking, not stand in for it. It’s satisfying to watch a profession of design experts land in the same place through daily practice.

Their advice to the rest of us is written between the lines of this study. Use the tool for the starting push. Keep the judgment for yourself. And treat the skill of using it well as something worth teaching on purpose.

References

Luo, T., Muljana, P. S., Ren, X., & Young, D. (2025). Exploring instructional designers’ utilization and perspectives on generative AI tools: A mixed methods study. Educational Technology Research and Development, 73, 741-766. https://doi.org/10.1007/s11423-024-10437-y

The AI in Education Research Digest

Keep exploring the research with me.

Get my reviews of new studies on AI, teaching, and learning, with thoughtful takeaways for your work as an educator.

Free to join. Unsubscribe anytime. Privacy policy.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top