How to Spot a Deepfake: A Practical Guide for Young People

I'm Parikshet. One of the most important AI skills of my generation is not making AI — it is recognising it. Deepfakes are getting better every year. By the time I finish secondary school, some deepfakes will be nearly impossible to detect visually. That makes knowing the detection process — not just the visual cues but the verification workflow — essential digital literacy.

What You Are Actually Looking For

Most deepfake detection guides give you a list of visual artefacts. That is a good start, but it misses the bigger picture. Your goal is not to identify a deepfake in three seconds of casual watching. Your goal is to establish whether a video should be trusted before you share it. These are different goals with different processes.

8 Visual Tells to Check

1. Blinking. Early deepfakes blinked incorrectly — too rarely, too perfectly rhythmically, or not at all. Newer models have improved, but look at blink rate: humans blink 15–20 times per minute in varied patterns.

2. Hair edges. Fine strands of hair are computationally hard to synthesise correctly. Look for blurring, flickering, or unnatural uniformity at the hairline.

3. Ear edges and jewellery. Earrings and ear contours often glitch in deepfakes — flickering slightly or appearing inconsistently rendered between frames.

4. Skin texture. Real skin has pores, fine lines, and slight texture variations. Deepfake skin often looks unnaturally smooth or plastic-like, especially around the cheeks.

5. Lighting inconsistency. Does the lighting on the face match the scene? A face lit from the left in a room lit from the right is a strong signal.

6. Lip sync. Play audio at 0.75x speed and watch lip movements carefully. Slight mismatches — audio slightly ahead or behind the visible mouth — are common deepfake artefacts.

7. Eye movement. The gaze does not always track naturally in deepfakes. The eyes may appear slightly glazed or the gaze direction may feel disconnected from the emotional content.

8. Background consistency. Objects near the face — hair in front of background, background elements near the face edge — may flicker or warp inconsistently as the face moves.

The Verification Workflow

Visual inspection is the last step, not the first. Here is the process I use:

Step 1: Context check. Where did this video come from? Is it from a verified account? A known news organisation? If it came from an anonymous account with no history, be sceptical.

Step 2: Source search. Search the person's name + the topic in Google News. If something this significant really happened, there should be multiple independent reports of it. If there are none, that is a major red flag.

Step 3: Reverse search. Take a screenshot from the video and run it through Google Images reverse search or TinEye. This can reveal whether the image has appeared in other contexts.

Step 4: InVID/WeVerify. This free browser extension analyses video metadata, reverse-searches video frames, and checks upload history. It was developed for professional fact-checkers but is free for anyone.

Step 5: Visual inspection. Only now look for the visual tells listed above. By this point, steps 1–4 will have resolved most cases.

Why This Matters More Than You Think

Deepfakes of students have been used for bullying. Deepfake audio has been used to impersonate parents calling schools. Political deepfakes have circulated during elections in multiple countries. If you share a deepfake without checking, you become part of the distribution network for misinformation — even if you had no idea it was fake. The five-step process above takes about two minutes. It is worth it before sharing anything you did not personally witness.