wunder beta🕵️ How to Spot a Deepfake
A computer can make a picture, a voice, or a video of something that never happened. The habit is not panic. It is how you check.
What you’ll learn
- Seeing Isn't Always BelievingUnderstand that a computer can generate a realistic picture, voice, or video of something that never happened, because it guesses what looks likely rather than checking what is true.A photograph used to be treated as strong proof. A deepfake is a picture, video, or sound a computer made or changed so it looks real. The program trains on huge numbers of real examples, then generates a best-guess image. It does not know whether the person or event is real.
- Clues You Can Sometimes SpotUse common visual giveaways as a starting look, and recognise that they are clues, not proof, because fakes improve and real photos can look odd.AI still images often slip on hands, teeth and eyes, writing, lighting, objects that should work, and waxy skin or warped backgrounds. Slow down and zoom in. Those tells are getting rarer, and an extra finger is not automatic proof, so looking alone cannot confirm an image is real.
- The Real Skill: Check the SourceTreat source-checking — who posted it, whether trusted places agree, and reverse image search — as the most reliable test.The dependable questions are where the image came from and whether other trusted sources report the same thing. Anyone can pause, find the original poster, search elsewhere, try a reverse image search, and ask a trusted adult.
- Pause Before You ShareTell harmless or labelled creative uses from deceptive ones, and pause before sharing so you do not spread a trick.AI media can be fun and fair when nobody is being tricked, or harmful when used to lie, scam, or bully. Sharing a fake, even by accident, helps it travel. Tonight, pause on anything that makes you go wait, really?, and check the source.
Questions this course answers
Why can't we fully rely on the old rule 'seeing is believing' any more?
AI can now generate realistic pictures, voices, and videos of events that never occurred, so an image by itself is no longer solid proof.
When an AI creates a new image, what is it mainly doing?
The program chooses what looks likely from patterns. It never checks whether the result is true.
Why do AI images sometimes get hands wrong?
Hands vary hugely and overlap in complicated ways, so pattern-based guessing struggles with them more than with common things like the centre of a face. A real person can have six fingers too, so an extra finger is a clue, not a verdict.
You study a photo closely and see nothing wrong. What does that prove?
The best fakes have no obvious mistakes, and real photos can look odd too, so not finding a flaw does not prove an image is real.
What is the most reliable question to ask about a surprising image?
Checking the source and whether reliable places agree is far more dependable than judging by appearance, which can be faked.
You see a shocking image from an account you've never heard of. What's the best first move?
Pausing to check the source before reacting stops you from being fooled and from accidentally spreading a fake.
Grounded in trusted sources
- MIT Media Lab, Detect DeepFakes project (no single tell-tale sign; video face-swap tells include skin texture, eyes and eyebrows, glasses glare, blinking, and lip sync — not a hand-and-text checklist), https://www.media.mit.edu/projects/detect-fakes/overview/
- Kamali, Groh, Nakamura, Chatzimparmpas, Hullman, How to Distinguish AI-Generated Images from Authentic Photographs, arXiv:2406.08651 (anatomical, stylistic, functional, and physics implausibilities; hands, teeth, eyes, text, lighting; artifacts are not always present; an extra finger is not automatic proof), https://arxiv.org/abs/2406.08651
- Matthew Groh et al., 5 Telltale Signs That a Photo Is AI-generated, Kellogg Insight, 9 Sep 2024 (diffusion models reverse noise to match a text description; they are not trained on spelling, physics, or anatomy; hands may lack nails, merge, or show odd proportions), https://insight.kellogg.northwestern.edu/article/ai-photos-identification
- Phil Swatton and Margaux Leblanc, What are deepfakes and how can we detect them?, The Alan Turing Institute, 7 Jun 2024 (deepfake = AI image, video, or audio made to mimic a person or scene; GANs vs diffusion; detectors struggle to keep up; the old no-blink tell was patched once people watched for it; not all deepfakes are meant to deceive), https://www.turing.ac.uk/blog/what-are-deepfakes-and-how-can-we-detect-them
- News Literacy Project, Tutorial: Reverse image search, 14 Aug 2025, and Tutorial: Lateral reading (leave the page; see what trusted places say about the source and the claim), https://newslit.org/news-and-research/reverse-image-search/ and https://newslit.org/news-and-research/lateral-reading/
- UK Safer Internet Centre, Misinformation (stories are often built to cause panic; think before you share; do not believe everything that looks genuine; compare with what you already know and look for support), https://saferinternet.org.uk/online-issue/misinformation
- Federal Trade Commission, Scammers use AI to enhance their family emergency schemes, 20 Mar 2023 (a short clip of a real voice is enough for a cloning program; do not trust the voice alone when someone who sounds like a relative asks for money), https://consumer.ftc.gov/consumer-alerts/2023/03/scammers-use-ai-enhance-their-family-emergency-schemes
Every Wunder lesson is built from real, reputable sources — never invented.
Related Science courses
Wunder is a personalized learn-anything platform — tell it any topic and it builds a beautiful, fact-checked course in minutes, with narration, a knowledge check, and a college-style University track.
Browse more Science courses · All topics · Home
© 2026 Wunder Learning LLC · Terms & Privacy