How to Spot a Deepfake Video or Voice Clone Before You Share It

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'll know it worked when 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'll know it worked when 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'll know it worked when you have a clear answer to those three questions — or a clear reason to be suspicious.

Run a quick source check

Take a screenshot of the person's face in the video (or use the platform's "save image" option). Open a reverse image search (commonly available on Google Images, TinEye or Yandex Images — no account needed) and upload the picture. Look for earlier appearances of the same face. If the face is supposed to be a public figure but the search shows nothing, or if older photos don't match the person in the clip, treat it as a likely fake. If the clip is supposedly a private person, send a quick message to that person on a channel you already had before the clip — not the one the clip came from — and ask if they really sent it.

You'll know it worked when you can either confirm the source or decide it isn't worth the risk.

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'll know it worked when you stop and think before tapping "forward."