AI tools can now create images, videos, and audio clips that look and sound convincingly real. These are often called "deepfakes" — synthetic media generated by artificial intelligence. While the technology improves every year, AI-generated content still leaves behind clues you can learn to spot.
This guide shows you what to look for across images, video, and audio — and what tools can help when your eyes and ears aren't enough.
How to Spot AI-Generated Images
Look closely at these areas when you suspect an image may be AI-generated:
- Hands and fingers: AI frequently struggles with hands. Look for extra fingers, missing knuckles, fingers that merge together, or hands that seem unnaturally positioned.
- Eyes and teeth: Zoom in. AI-generated faces may have pupils that are different sizes, irises with unnatural patterns, or teeth that look too uniform — almost like a single white block rather than individual teeth.
- Skin and textures: AI-generated skin can appear too smooth, almost plastic-like, without the natural pores, blemishes, and texture variation found in real photographs. Backgrounds and fabrics may also look oddly blurry or too perfectly patterned.
- Text and lettering: AI has difficulty rendering readable text within images. Look for garbled letters on signs, books, or clothing — words that almost look right but don't quite make sense.
- Edges and borders: Check where the subject meets the background. AI often produces blurring, smudging, or unnatural halos around hair, jewelry, and clothing edges.
- Symmetry and consistency: Look at earrings, glasses, shirt collars, and other accessories. AI may render the left and right sides slightly differently.
How to Spot AI-Generated Video
Deepfake videos add a time dimension to all the image problems above, plus a few more:
- Lip sync: Does the speaker's mouth movement match the audio precisely? Even small mismatches — slightly delayed lip movement or lips not fully forming certain sounds — can indicate AI manipulation.
- Blinking and facial expressions: AI-generated faces sometimes blink too rarely, too regularly, or not at all. Micro-expressions (brief, involuntary facial movements) may be absent or look mechanical.
- Head and body movement: Watch for unnatural rigidity. Real people shift their weight, tilt their head unevenly, and move in slightly unpredictable ways. AI-generated figures may look "locked in place" from the neck down while only the face moves.
- Lighting inconsistencies: Shadows on the face should match the lighting of the environment. If the face appears to be lit from a different direction than the rest of the scene, the video may be manipulated.
- Edge flickering: Around the jawline, hairline, and ears, look for a brief "shimmer" or flickering effect — a frame-by-frame instability that is a common artifact of face-swapping technology.
How to Spot AI-Generated Audio
Voice cloning has become remarkably convincing. Here's what to listen for:
- Flat or unnatural rhythm: Real speech has natural hesitations, breaths, and changes in pace. AI-generated speech may sound too smooth, too evenly paced, or lack the small imperfections of genuine conversation.
- Unusual pauses or breathing: Listen for breathing patterns. AI audio may have no audible breathing between sentences, or breathing that sounds artificial and inserted at regular intervals.
- Robotic undertone: Even high-quality cloned voices can have a subtle metallic or synthetic quality, especially on certain syllables or when the speaker raises their voice.
- Emotional mismatch: If the words convey urgency or distress but the tone sounds flat or detached, this may indicate a synthetic voice. Real emotion creates natural variation in pitch and speed that AI often fails to replicate convincingly.
Use Detection Tools When In Doubt
Your eyes and ears are a good first line of defense, but they're not always enough — especially as AI improves. These tools can help:
- Trend Micro ScamCheck — Scans links, messages, and media for signs of scams and AI manipulation. Available as an app and browser extension.
- Reverse image search — Upload a suspicious image to Google Images to see if it appears elsewhere or if similar versions exist.
- Metadata check — Right-click an image and check its properties. AI-generated images sometimes lack camera metadata (EXIF data) or contain unusual software tags.
🔔 Important: No single detection method is 100% reliable. If something feels off — a surprising video of a public figure, a voice message demanding urgent action, or a "too perfect" photo — treat it with skepticism and verify through a separate channel before acting on it.
What To Do If You Find a Deepfake
- Don't share it. Sharing deepfakes — even to debunk them — can amplify their spread.
- Report it to the platform where you found it. Most social media platforms have specific reporting options for manipulated media.
- Save evidence. Take a screenshot or screen recording before the content is removed, in case you need it for a report later.
- Warn the person depicted, if you can identify them. They may not know the content exists.