What we learned from making AI-generated videos for teachers
It's becoming harder to tell the difference between authentic video and AI-generated content. What started as a novelty on social media is increasingly showing up in educational resources, professional learning, classroom materials, and, least surprising, in blockbuster movies. AI-generated media has become easier to produce and harder to detect. That’s why we’re writing this post. In education, we think transparency is paramount when it comes to using AI production and tools. Teachers deserve to know what goes into the materials they use in class.
Over the past two years, we’ve developed more than 30 videos with human writers, production houses, and animators for the OER Project website, YouTube, and social channels. Humans also design, write, and produce the activities, articles, comics, and other videos in our course. This summer, we began experimenting with AI-generated videos for teacher professional learning. This isn’t a shift in our approach. Rather, it’s an experiment. We wanted to know whether we could use AI videos in limited ways to help teachers grapple with ideas that are hard—and often prohibitively expensive—to represent via traditional media production.
Early on in this experiment, we realized that we’re grappling with the same question you are: Should AI-generated videos have a place in the classroom? As with most questions involving technology, the answer is simple: It depends.
Our exploration of this question led us to a simple decision framework teachers can use: Start with the learning goal, scrutinize the content, be transparent with students, and rely on human judgment. The most important question isn’t whether a video was created with AI. It’s whether the video helps students understand something more accurately, clearly, and meaningfully.
Our case study: sourcing videos for teachers
To better understand the ins and outs of AI videos, we felt that it was necessary to create some ourselves so that we could investigate the process from end to end, including how a teacher and or a student might interact with the video that we made. We knew AI video would never be the right answer for topics for which rich historical imagery already exists; no generated image can replace the real thing. So we experimented with teacher professional learning videos about sourcing—videos about how teachers can help students pause, examine where information comes from, and question the creator before accepting it as true. We tried a few approaches. Some attempts gave us cool-looking videos, but they didn’t have the instructional oomph we wanted. That trial-and-error process became our case study. We learned that AI-generated videos are only useful when they help solve a real teaching and learning problem.
We wanted to support teachers in making sourcing feel relevant in a world where misinformation is increasingly difficult to identify. We set out to provide a series of quick teaching tips and ideas that teachers could implement with minimum prep. From there, we developed scripts, reviewed visuals, checked claims, and asked whether each video would actually help teachers model the kind of thinking required of students. In other words, AI helped create the videos, but it didn’t set the goals. It didn’t determine the messages we deliver. And it didn’t decide whether the final product was classroom worthy. Humans did.
That experiment left us with a few practical benchmarks teachers can use before bringing AI-generated video into the classroom.
Start with the learning goal
Before evaluating the technology behind a video, ask the same question you’d ask of any classroom resource: What do I want students to learn from this? Is the video helping students visualize a process? Introducing a concept? Building background knowledge? Sparking discussion?
A strong video, whether AI generated or not, has a clear instructional purpose. For our sourcing videos, that purpose was to help teachers introduce durable classroom habits. Before students trust information, they should get used to asking a series of questions. Who created this? Why did they make it? What evidence supports the information in it?
AI-generated videos work best when they fill a visual gap. If authentic footage, primary sources, historical photographs, or a well-produced documentary clip already exists, those resources are usually the better choice. For our sourcing videos, we weren’t trying to re-create a historical event or replace primary evidence. Instead, we wanted to make an abstract thinking process visible for teachers so they could model that process for students.
Watch out for “confidently wrong”
One big challenge with AI-generated media is that it’s gotten so good, even its errors can look convincing. A video may include inaccurate visuals, incorrect pronunciations, misleading imagery, or factual mistakes. But because the content looks polished, students may assume it’s trustworthy. For example, AI-generated historical imagery might show the 1940s in crystal-clear, full-color detail. It might look more real to students than a grainy—but genuine—photograph. Even if the scene is based on a real photograph, that level of polish can distort students’ understanding of what authentic historical evidence from that time looks like. Students need to learn the constantly shifting ways to identify AI forgeries. This is why when you are evaluating AI-generated video, we encourage you to look for the following before bringing it into your lesson:
- Historical inaccuracies
- Scientific errors
- Mispronounced names or terminology
- Misspellings or incorrect labels
- Visuals that don’t quite match the narration
- Imagery that exaggerates or distorts the topic
- Colors, styles, or backgrounds that don’t align with the period
Treat an AI-generated video the way you would any source. When we created our videos, that meant reviewing every element before anything was ready for teachers to view. In the classroom, it means previewing with the same care you would bring to a primary source, article, or documentary clip.
Pay attention to tone
Accuracy isn't the only thing that matters. How information is presented can shape interpretation. In our first round of sourcing videos, we tried to lean into the investigative element of sourcing, and we chose an archaeological theme. However, the videos came across as cartoonish and silly in a way that made the visual theme the focus, rather than the pedagogical tips we wanted to convey.
That experience reminded us that a video can include accurate information and still send the wrong instructional message. In the second round, we paid closer attention to whether the visuals, pacing, and overall approach supported the habits we wanted teachers to model: slow down, look closely, and consider what evidence is actually there.
Be transparent with students
We believe students should know when they’re viewing AI-generated content. As educators, we teach students to evaluate sources, identify bias, and question how information is created. AI-generated media offers another opportunity to develop those skills.
A simple acknowledgment that a video contains AI-generated visuals or narration can lead to valuable conversations about media literacy, misinformation, and digital citizenship. In our own work, we state clearly when AI was used to help produce a resource, and we explain why that choice supported the learning goal. You don’t need to make AI the whole lesson, but transparency can help your students become more thoughtful consumers of media.
You might ask students:
- How can we tell whether something is AI generated?
- Does knowing a video was AI generated change how we evaluate it?
- What evidence would we need to verify the claims being made?
- Could someone mistake this content for authentic footage?
These conversations are becoming increasingly important as AI-generated content becomes more common in students’ everyday media lives.
Human judgment still matters
AI can help create a video, but it can’t decide whether that video belongs in your classroom. That still requires educator judgment. Our AI-generated videos were scripted, reviewed, and fact-checked by people who understand the content and the teachers who will use it. But don’t take our word for it. In any video you use, look for evidence that knowledgeable humans shaped and reviewed the resource.
Then, if the video passes your analysis and you want to share it with students, help them analyze the media as they would any other source—by activating their claim testers:
- Is the evidence in the video historically accurate? Are there other sources you’ve reviewed that can corroborate what the video shows?
- Do you feel like this video accurately reflects the time period, person, or event being shared?
- Who created the video and are they a trustworthy source? How do you know?
- Based on what you know about history, does this video support, extend, or challenge your understanding?
While this analysis works for AI-generated video, it can extend to any media you choose to bring into your classroom.
The right tool for the job
In our experiment with these videos, we learned that AI can be valuable in certain scenarios. It can help make invisible thinking visible, and it can help fill a gap where a stronger resource doesn’t exist. But our standards for curriculum quality shouldn’t change. Don’t use AI just because it’s the shiny new thing or because it feels easier. Only use it when it helps students learn something more clearly, accurately, or thoughtfully.
AI-generated video is a tool. Like any tool, its quality is dependent on the skill of the wielder. AI can’t replace teacher judgment. If a video helps your students see, question, or understand something better, it may deserve a place in the lesson. If it fails that test, you should feel confident leaving it out.