This post was coauthored by Lydia Rowley, a student at Brigham Young University.
Imagine this: You unlock your phone and open Instagram. The first post you see is a video of a popular public figure making rude, racist comments. You’re stunned. You never thought they would endorse those viewpoints. As your finger hovers over the unfollow button, you watch the video again—and notice a few little details that change everything. The hair and skin are a little too shiny. The motions are too smooth. A few people in the background disappear between shots. Clicking on the comments confirms your suspicions: This video is AI-generated.
You close Instagram, relieved to realize the truth. But the fear lingers: Would you notice next time? Even more concerning, would your children notice?
Visual Deception
Generative AI becomes more realistic with every update, making it ever more difficult to differentiate AI-generated from real content. Some studies claim children cannot differentiate AI from real content as well as adults, who are not particularly successful themselves. However, new data from our under-review paper shows that 9-year-old children did better than guessing when asked to identify AI-generated videos, though less than half (46.5 percent) identified all the videos correctly. Children’s understanding of AI is related to their ability to identify it. The 54.7 percent of children in our study who said they knew what AI was correctly identified more videos than their less familiar peers.
Parents are often a child’s first source of information, so in-home education about AI-generated content is essential. But what do we teach children about how to identify this new technology? In our research, currently under review, children most readily identified an AI-generated video involving babies riding animals in a fashion show, explaining that this could never happen in real life. Some children noticed when the AI made errors with lighting, skin, or hair, but surprisingly few made note of a caption stating that the video was AI-generated. These observations show us what we can do to help children learn to identify AI-generated videos.
Teaching Children to Identify AI
How can we teach children to identify AI-generated videos? The following suggestions build on our findings.
- Start early. The children in our study were 8 and 9 years old, but most understood our explanation of AI, and many were already familiar with it. While relatively few could give an accurate definition, children of this age are ready to learn about AI and can understand what it is and what it does.
- Be concise and use examples. Children can understand what AI is, but they are more likely to retain and understand a simple, age-appropriate definition with relevant examples. For instance, parents could explain that AI is a computer that is programmed to think and learn like a person. Parents can share that AI is made to answer questions, solve problems, recognize patterns, and generate text, images, audio, and videos. Relevant examples of AI include Siri, Google, Alexa, Netflix’s recommendation system, and chatbots like ChatGPT. Including the names of specific image generators (Sora, Seedance, etc.) and the logos associated with them can help children notice and understand the watermarks that appear on many AI-generated images and videos.
- Start with a reality check. The most common reason children gave for saying a video was AI-generated was that it couldn’t happen in real life, so encouraging a reality check is a good place to start. Prompt your children to consider if the content could occur in real life, or if the content looks, sounds, or behaves like it should in real life. Be sure to address the difference between acting in a movie, which is not real life but is done by real people, and AI-generated videos.
- Talk about “tells.” Some children identified specific AI-related digital artifacts, or “tells,” in the videos. Pointing these out helps children know what to look for in the future, so instruct children to look for "tells" like nonsense words, anatomy problems, odd lighting, blurriness, oddly smooth surfaces, mismatch between voices and mouth movements, and inconsistency in objects, background, or motion.
Conclusion
As AI-generated videos and images become increasingly common on the internet, parents can prepare their children to thoughtfully and critically evaluate this content. By educating children with age-appropriate definitions and relevant examples, discussing the implausibility of some AI-generated content, and explaining specific “tells” to look for while evaluating online content, parents can help children develop the critical thinking skills necessary to differentiate real and AI-generated content. AI-generated content is not going away anytime soon, but by educating their children, parents don’t need to be quite so afraid of its deception.