BeginnerLesson 2 of 59 min read
Doctored image or video: investigate before you believe
A course from the Academy: Scams, deepfakes and cloned voices: staying safe
Go to the quiz ↓A photo of an explosion doing the rounds on social media, a video of a politician saying something shocking, a clip from a celebrity interview. Before sharing, and above all before acting, one question: is it real? Visual clues help, but they have their limits. The real method is to investigate.
The clues you can still look at
- In a video: lips that don't quite follow the words, odd blinking, blurry edges around the face when the head turns, glasses or jewellery that warp, a voice that's too even, with no breathing.
- In an image: strange hands or ears, shadows that don't point the same way, a background that warps, skin that's too smooth, lighting that's too perfect.
- In the context: a very sharp image of an event no news outlet is talking about, a video cut off with no beginning or end, an unknown account posting it "exclusively".
The limits of these clues
Each new generation of tools fixes the flaws of the last one: a clue that's visible today may be gone in a few months. The reverse is true too: a real video, compressed, filmed at night or in poor quality, can show flaws that look like those of a fake. A clue raises your suspicion; on its own, it proves nothing. Even automatic detectors get it wrong, in both directions.
Investigating in three moves
- Go back to the original source: who posted this image first? A recognised news outlet, an official account, or an anonymous account created three weeks ago? An image circulating with no author, no place and no date should be treated as unverified.
- Run a reverse image search: instead of typing words, you give the image itself to a search engine (Google Lens, TinEye or Bing, for example). It shows you where the image has already appeared, and since when. A "photo from yesterday" that was posted elsewhere years ago: case closed. For a video, take a screenshot of a sharp moment and search for it the same way.
- Cross-check with reliable media: if the event is real and important, several newsrooms will cover it. The fact-checking sections of major news outlets often deal with images that go viral.
Some tools add an invisible watermark or a provenance label to the images they generate. When it's there, it's a great help; when it's missing, that proves nothing.
Key takeaways
- Visual clues raise suspicion; they prove nothing.
- They go out of date: each generation of tools fixes its flaws.
- Go back to the source, run a reverse image search, cross-check with reliable media.
- When in doubt, don't share, and don't act.
Key concepts
Complete the lesson
A 4-question quiz: 3 correct answers are enough, plus an exercise to practise. It opens in the Academy, on this same lesson.
Take the quiz and complete the lesson Free. With an account, your progress is saved and leads to the course certificate.The course: Scams, deepfakes and cloned voices: staying safe
- The new scams AI makes possible9 min
- Doctored image or video: investigate before you believe9 min
- Cloned voices: hang up, call back, family password9 min
- Protecting your loved ones, and what to do if you're targeted10 min
- The 4-step reflex, with a case study10 min