AI WORLD HQHow to Spot a Deepfake Video or Voice Clone Before You Share It
Five quick checks anyone can run on a suspicious video or voice clip in under two minutes — no app or account needed
This guide was written by AI. It passed automated fact and quality checks; no human editor reviewed it.
A short clip of a public figure saying something outrageous lands in your family group chat. A frantic voicemail from your "boss" asks for gift cards. None of these are obviously fake, and that's exactly the point. A deepfake is a video, image or audio clip made with AI to look or sound like a real person who never actually said or did the thing, and voice cloning (reproducing someone's voice convincingly from a short audio sample) is now easy and cheap. The good news: a few careful seconds of looking is usually enough to catch the most common giveaways. Even so, no single check is perfect, so your final safety net is always verifying with the real person before you share.
💡 Tip: tap a step’s number when you finish it — a green tick appears and your browser remembers how far you got.
- The clip itself (saved to your phone or laptop, or a link in a message)
- A web browser (for the reverse image check in Step 4)
- About 2–3 minutes per clip
- No account, no app, no paid plan needed
Watch the face for small glitches
Pause the video and study the face for about 10 seconds. Focus on the eyes, the edges of the face, and the mouth. Common giveaways: eyes that don't quite blink in time; glasses that seem to slide or warp when the head turns; teeth that blur into a single white block when the mouth moves; a face outline that looks soft or "painted" where it meets the hair and background; lighting on the face that doesn't match the rest of the scene. If you can't pause or zoom, slow playback to 0.5x and look again. On a phone, tap the screen to bring up the playback speed option — it's usually a small "1x" button near the timeline.
you spot at least one thing that feels off — or, just as usefully, confirm the face looks natural.
Listen to the voice for flat delivery
Close your eyes for a moment and just listen. Real human speech has small imperfections: a quick breath, a half-stutter, a sentence that gets quieter when the speaker is unsure. AI-generated voices (TTS, short for text-to-speech — software that turns written words into spoken audio) tend to be too clean, too even in volume, and oddly flat in emotion. Listen for sentences with no breath at all, words that all sit on the same pitch, or a tone that doesn't match the situation (someone supposedly panicking, but the voice sounds calm and steady). On a phone, plug in headphones so background hiss doesn't mask the small tells.
you can tell whether the voice feels human or synthesized (made by a computer, not a real person).
Read the message around the clip
Before you forward anything, stop and read the context. Who sent it? Is it really them, or just a forwarded link from someone you trust? What does the clip ask you to do — donate money, click a link, send a code, stay quiet about something? Most scams lean on urgency, secrecy or big emotional claims: "send this in the next hour", "don't tell anyone", "this is huge." Those are the same red flags as any scam, and they're a strong reason to slow down. If you can't answer "who, what, and why" from the message alone, that's already a signal.
you have a clear answer to those three questions — or a clear reason to be suspicious.
Decide what to do next
If any of the checks above raised doubt, the safest move is simple: don't share it. Reposting a fake — even with a "this might be fake" caveat — still spreads the original. Tell the person who sent it, politely, that you couldn't verify it, and ask them where they got it. If the clip claims to be a real person you know, give them a heads-up that someone may be using their face or voice. If the clip is a scam attempt, report it to the platform where you saw it, and if money or account access was involved, contact your bank or the company being impersonated directly.
you stop and think before tapping "forward."
- Trusting the "share if you agree" framing. If a clip is trying to get you to forward it before you've checked, that's the scam working. Slow down — the share button can wait 30 seconds.
- Assuming a friend's name means a friend sent it. Accounts get hacked and contacts get spoofed (a message made to look like it came from someone you know). Verify through a separate channel before trusting.
- Relying on a single "AI detection" app. No consumer app is reliable enough to be your only check, and some detection tools themselves are scams. Your eyes, ears and a quick source check beat any single app.
Pick one short video you saw recently that made you feel a strong reaction. Spend two minutes running Steps 1, 2 and 4 on it. Even if it turns out to be real, the practice makes the next check faster and easier.
❓ Quick questions
How long does this take?
About 6 minutes — the guide has 5 steps, and you can tick each one off as you go.
Do I need to prepare anything?
- The clip itself (saved to your phone or laptop, or a link in a message)
- A web browser (for the reverse image check in Step 4)
- About 2–3 minutes per clip
- No account, no app, no paid plan needed
What mistakes should I avoid?
- Trusting the "share if you agree" framing. If a clip is trying to get you to forward it before you've checked, that's the scam working. Slow down — the share button can wait 30 seconds.
- Assuming a friend's name means a friend sent it. Accounts get hacked and contacts get spoofed (a message made to look like it came from someone you know). Verify through a separate channel before trusting.
- Relying on a single "AI detection" app. No consumer app is reliable enough to be your only check, and some detection tools themselves are scams. Your eyes, ears and a quick source check beat any single app.
Keep reading
✦ Generated by AI in AI World HQ's automated newsroom, written in plain language. Checked by automated fact and quality gates — no human editor reviewed this guide. Spot a mistake? Use the buttons above.
📘 Prefer it in one file? Scams, Deepfakes and Privacy: The AI Safety Pack (9 guides, free printable PDF)
← Back to all stories