As students head back to school, AI-generated images and videos are adding another complication to figuring out what to trust online. A convincing clip can show something that never happened, and an authentic photograph can still mislead if someone changes its caption. Deepfakes, which use AI to imitate a person's appearance or voice, make the question harder when the person seems familiar. Libraries have several practical ways to help students check what they're seeing, from finding an original recording and comparing sources to understanding what an AI detector can actually tell them. These are skills you can work into a class visit, a media literacy program, or a conversation at the reference desk.
Check the Source
Before you start counting fingers in a photograph or replaying a strange facial expression, find out where the post came from. Who first shared it, when did it appear, and what did the original caption say? An old photograph with a new caption can spread a false claim without anyone using AI at all.
"One of the most important lessons we can teach students is that there is no single visual clue that reliably proves something was generated by AI," says Halil Ibrahim Dursunoglu, a faculty specialist in computer science at Western Michigan University whose work includes cybersecurity and deepfake detection.
A reverse image search using Google Lens is a useful place to start. Upload the image or paste its image address to look for other appearances, and compare the dates and descriptions. For a video, try searching screenshots from a few different moments. You might discover that a clip labeled as today's storm actually shows an unrelated event from years earlier.
Dursunoglu also recommends searching for the event or statement itself. If a clip appears to show a school official announcing a new policy, look for the full recording or a dated announcement on the district's website. Check independent reporting, too, and look at how the information was confirmed. Repeated reposts may all point back to the same unverified clip. If you can't find confirmation, the claim remains unverified; that alone doesn't establish that it's false, especially for a new or local event.
Pause Before Sharing
It's worth paying attention to how a post makes you feel, particularly when you're tempted to pass it along immediately. Ticki Favaroth, founder and CEO of VORIX, whose work includes digital identity and credential verification, offers straightforward advice: "If content creates an immediate emotional reaction, don't repost it right away. Investigate it."
That doesn't mean an upsetting story is false. It means the moment you feel compelled to share it is a good time to check what you're sharing. Ask students what the post wants them to believe or do. Is it reporting an event, selling something, making a joke, or asking viewers to spread an accusation? An openly labeled AI illustration serves a different purpose from a fabricated recording presented as evidence. Identifying the actual claim gives students something concrete to investigate, even when they can't tell how the image was made.
Look Closely
There is still value in slowing down and examining the footage. Sergey Rodin, general producer at Lava Media, works with both filmed and AI-generated video and recommends watching the background for objects that change or disappear unexpectedly. To catch these details, he says, "you have to pause, scrub back and forth, and look past whatever the shot wants you to look at."
This is easy to demonstrate during a library session. Replay a short clip and have students follow one background object through the scene. Does it stay consistent, or does it change during a camera movement or an ordinary edit? For a clip of someone speaking, listen while watching their mouth to see whether the movements appear to match the words.
Be careful about turning these observations into proof. Dursunoglu notes that poor lighting, editing, compression, and ordinary camera artifacts can create suspicious-looking effects in genuine recordings. A strange frame is a reason to investigate, while a clip that looks flawless still needs a source check. The same goes for audio: if a familiar voice says something surprising, look for the original recording and confirmation of the statement.
AI Detection Tools
An AI detector may look like an easy answer, but its results need a closer look, too. These tools can miss generated material or flag authentic content, and a tool built to check writing shouldn't be assumed to work on images, videos, or voices. "New generation techniques, compression, cropping, screenshots, filters, and re-encoding can all affect detection accuracy," Dursunoglu says.
The National Institute of Standards and Technology's review explains how performance varies with the type of media, the generator, and changes made to the file. If you use a detector with students, talk through what it measures and what its result leaves unanswered. A score shouldn't become the sole basis for declaring a video genuine or accusing a student of fabrication.
You may also see Content Credentials, records that can show how a file was created or edited, including certain uses of AI. The Coalition for Content Provenance and Authenticity explains how these work. They help establish a file's history and whether that information has been tampered with, but they don't certify that the story attached to it is true. Missing credentials don't prove that something is fake, either. Participation is optional, and those records can become separated from a file when it's shared.
Practice in the Library
You don't need an expensive program to give students practice with these skills. Victor Osondu, founder of the UK AI education practice AI Tutorium, suggests having secondary students create an AI image or short clip and discuss what makes it convincing. "I tell my trainees that AI wisdom is mostly about asking better questions," he says.
You can adapt that idea into a comparison activity using a genuine photograph with its original caption, another genuine photograph with a misleading caption, and an AI-generated image. Keep an answer key with the source and history of each example. If students are generating media, use a school-approved tool and fictional subjects. Preparing the examples yourself also lets everyone take part without needing individual accounts.
Have small groups investigate the posts and explain what they found. They should be able to identify the claim, show the sources they checked, and say what remains uncertain. Dursunoglu suggests asking, "What evidence would you need before you would feel comfortable sharing this as authentic?" Give students credit for showing their reasoning and revising an initial impression as they learn more.
There are resources you can use without building a lesson from scratch. The News Literacy Project offers AI literacy materials, and its reverse image search tutorial can support a demonstration. A few minutes spent working through a questionable post together can make source checking easier to understand.
Why It Matters
Students need room to say they don't know yet, especially when the available evidence doesn't support a firm answer. Libraries can make that a normal part of evaluating information by showing how to follow up and where to look next. Try bringing one of these checks into your next class visit or media literacy program, and let students talk through what changed their minds. When they leave, they should have a checking habit they can use and a place to ask for help when they're still unsure.
Enjoyed this article? Subscribe to Video Librarian today for access to over 40,000 pages of film resources tailored for librarians, educators, and non-theatrical audiences.