How to Spot AI-Generated Images and Deepfakes in 2026
AI images are now good enough to fool most people at a glance. These checks help you slow down and verify before you share.
Realistic AI-generated photos, videos and voices are now cheap and fast to make. Most are harmless fun, but they're also used in scams, fake news and impersonation. You don't need special software to protect yourself — mostly you need a habit of pausing and checking.
Start with the context, not the pixels
Visual tells are getting harder to see as the technology improves. The most reliable first questions are about where the content came from:
- Who posted it first? A brand-new or anonymous account sharing "breaking" footage is a red flag.
- Is anyone reputable reporting it? Major real events are covered by multiple established news outlets within hours.
- Does it make you feel a strong emotion? Outrage and urgency are what make fakes spread. That feeling is your cue to slow down.
Visual clues in images
- Text and signs: lettering on shop signs, labels or clothing that's garbled or nonsensical.
- Hands, teeth, ears and jewellery: extra or merged fingers, mismatched earrings, odd tooth shapes.
- Backgrounds: people in crowds with smeared faces, railings or windows that don't line up.
- Too-perfect lighting: glossy, airbrushed skin and dramatic light in a supposedly casual phone photo.
None of these prove an image is fake, and good fakes may show none of them. Use them as reasons to investigate further.
Deepfake video and cloned voices
- Lip movements that drift out of sync with speech.
- Unnatural blinking, stiff neck or shoulders while the face moves.
- A voice that sounds right but has flat emotion or odd pauses.
Voice cloning is increasingly used in "family emergency" scams. A caller sounds like a relative and urgently asks for money.
Free ways to verify
- Reverse image search: use Google Lens, Bing Visual Search or TinEye to find where an image first appeared.
- Check for content credentials: some cameras, apps and platforms now attach provenance labels showing whether an image was made or edited with AI. Look for an "AI-generated" or content-credentials label where platforms display one.
- Look for fact-checks: search the claim plus the words "fact check".
- Zoom in: view the full-resolution version; many artefacts vanish in thumbnails.
The 10-second rule before you share
Before reposting anything surprising: who made it, who else is reporting it, and would it matter if it were fake? If you can't answer the first two, don't share it yet.
