Jeremy Carrasco makes videos under the handle @showtoolsai on TikTok advocating for AI video literacy and pointing out that you can see if the videos in your feed are AI-generated.
AILSA CHANG, HOST:
AI is everywhere, right? Like, you're probably seeing this all over your social media feeds. Remember those cute bunnies jumping on the trampoline? False. And now, since there are so many of these fake videos, a kind of video critics with artificial intelligence have appeared.
(SOUNDBITE OF ARCHIVED RECORDING)
JEREMY CARRASCO: This video has a different AI model. I think it's Kling, a Chinese model, because the writing on the ceiling is in Chinese. And when he runs his hand through his hair, his little finger goes all the way up. They sell research chemicals and skin care products. This can't be true.
CHANG: This is Jeremy Carrasco. He's a video producer who spent years checking real videos for errors. Now he has turned his attention to artificial intelligence. And on TikTok and Instagram, he teaches people to recognize fakes. Carrasco says spotting counterfeits has become much more difficult as artificial intelligence models have improved. Missing limbs and fingers is no longer a small matter. But the models are still not perfect.
CARRASCO: For example, there are AI scams where people sit in hospitals and ask for money. But if you send these images to a doctor, he will realize that the equipment is completely wrong or half made. If you watch a video of a rock climber, the ropes will be unsafe or not tied to anything.
CHANG: (Laughter).
CARRASCO: And as a casual observer, you may not know that. But the good news is that you have other superpowers or other domain expertise that may come in handy in other situations.
CHANG: So, even if the scene looks realistic, your own logic may tell you it's fake. He says even something like camera angles can help you think critically about what you see.
CARRASCO: Absolutely. The main difference between these videos and our real ones is that there is no real person behind the camera. Therefore, in many cases, camera movement does not make sense. And to do this, you just need to take a step back and ask: who would direct this and why? There have been many viral videos of dogs rescuing babies from falling shelves where you had the camera positioned perfectly.
CHANG: Yes.
CARRASCO: You know, in real life this wouldn't happen. And if it looked like someone was holding a cell phone and a child was plugging something into the socket, that person would put the film down to help, you know?
CHANG: (Laughter) Right, right. Or like the shot where a cat jumps on someone's bed in the middle of the night and snatches a snake from it. Like, why would you point a security camera at yourself while you sleep at night?
CARRASCO: Exactly. I mean, not all of them are that obvious. But I'll tell you, if you can grasp these obvious things: pet videos are great because pet videos are popping up all over our feeds. If you can catch these obvious things and learn from them in other details, then you are well on your way to becoming better at recognizing them in general.
CHANG: Well, I saw that TikTok has a policy that requires creators to label all AI-generated content that contains realistic images, audio and video. This is politics. How reliable is this policy? And how common are these policies among social media companies these days?
CARRASCO: Very unreliable. And they have some detection. But detection appears to be based more on descriptions or hashtags rather than actual video analysis. As far as I can tell, no social media platform takes detection or labeling seriously.
CHANG: Okay. Well, that brings me to my next question. Some of your videos talk about technical watermarks that AI generators embed into their content. But are these indicators reliable? I mean, how are these watermarks even visible to the average consumer who encounters these videos?
CARRASCO: Yes, there are visible and invisible watermarks. That's why artificial intelligence companies like Google are trying to introduce invisible watermarks embedded into pixels. But if you give Google a photo that was created by AI but not Google, they won't be able to tell you if it's AI.
CHANG: Right.
CARRASCO: So, there is no good, reliable, centralized solution right now. But there are definitely some promising ones that hopefully we'll see come to fruition soon.
CHANG: I mean, God, talking to you right now and thinking about all these techniques that you can use to detect fakes, you know what I feel like I'm getting lost here is just the spontaneous delight that you would otherwise feel when you come across a funny video.
CARRASCO: Yes.
CHANG: And I'm wondering, how do you think this hurts all those video creators who are creating the real thing?
CARRASCO: I agree with you. This was the saddest part, because actually the best way to learn to recognize AI videos is to experience that feeling of wonder that you may have felt…
CHANG: Yes.
CARRASCO: …And you have to ask, wait, is it because it's AI now?
CHANG: Yes.
CARRASCO: And I would say that there are a lot of true creators who still give you that feeling of excitement. And you can just follow them and know that you can be sure that they won't scam you because a lot of real creators really aren't interested in making AI videos. And AI accounts are really just there to create AI. So I don't think this sense of wonder should go away. You just need to know where to look for it and not rely on social media algorithms to simply present it to you.
CHANG: This is AI literacy teacher Jeremy Carrasco. Thank you very much for your ideas.
CARRASCO: Thank you, Ailsa.
© 2025 NPR. All rights reserved. Visit our website terms of Use And permissions pages in www.npr.org for more information.
The accuracy and availability of NPR transcripts may vary. Transcript may be modified to correct errors or accommodate audio updates. Audio on npr.org may be edited after it is originally broadcast or published. The authoritative recording of NPR programs is the audio recording.
