Deepfakes in 2026: How to Spot AI-Generated Videos, Audio, and Images

In February 2024, an employee at a multinational engineering firm in Hong Kong transferred $25 million after a video call with what appeared to be the company's chief financial officer and several other colleagues. Every person on that call was a deepfake. The voices were synthetic. The faces were generated in real time. The employee was the only real human in the meeting.

That story sounds like science fiction. It happened.

And the technology has only gotten better — and more accessible — since then.

In 2026, deepfake technology has reached a point where casual detection by the human eye and ear is no longer reliable. We need new skills, new tools, and a fundamentally different relationship with digital media. This article is about building all three.

What Deepfakes Are (and How They've Evolved)

A deepfake is AI-generated or AI-manipulated media — video, audio, or images — designed to convincingly depict something that never actually happened. The term originally referred specifically to face-swapping in video, but it now covers the full spectrum.

Video deepfakes can put any face onto any body in video footage. They can make politicians appear to say things they never said, create fake pornographic content using real people's faces, or fabricate entire video conversations.

Audio deepfakes (voice cloning) can replicate any person's voice from just a few seconds of sample audio. The cloned voice can then say anything the creator wants. We covered voice cloning scams in detail in our article on AI voice scams.

Image deepfakes can generate photorealistic images of people who don't exist, place real people in situations they were never in, or create fake documents and screenshots that are indistinguishable from real ones.

What's changed in 2026 isn't just the quality — it's the accessibility. Tools that required significant technical knowledge two years ago can now be used by anyone with a laptop. Open-source models are freely available. Cloud-based deepfake services exist. The barrier to creating convincing fakes has collapsed.

How Deepfakes Are Being Used Against Regular People

You might think deepfakes are a problem for politicians and celebrities. They're not. Increasingly, they target ordinary people.

Financial fraud. The Hong Kong case is the most dramatic example, but smaller-scale versions happen regularly. A deepfake voice call from "your boss" asking you to wire money or share credentials. A deepfake video call from "your bank" asking you to verify your identity. These work because they exploit trust — you believe what you see and hear.

Romance and sextortion scams. Scammers create deepfake images or videos using victims' social media photos and threaten to distribute fake explicit content unless a ransom is paid. The content is entirely fabricated, but the emotional impact is real and devastating.

Misinformation and election interference. Fake videos of political figures making controversial statements can go viral before fact-checkers can respond. Even after debunking, the damage is done — the emotional impact of seeing and hearing a person "say" something persists even when you know it's fake.

Identity fraud. Deepfake technology is being used to bypass identity verification systems that rely on video selfies or face matching. Some criminals have successfully opened bank accounts and cryptocurrency wallets using deepfake identities.

How to Spot Deepfakes

I have to be upfront: spotting high-quality deepfakes is becoming genuinely difficult. But there are still indicators, and training yourself to look for them gives you a significant advantage.

Video Deepfakes: What to Look For

Unnatural eye movement and blinking. Deepfake models have historically struggled with realistic eye behavior. Watch for eyes that don't move naturally, staring that seems too fixed, or blinking that seems mechanical.

Inconsistent lighting on the face. The lighting on a deepfaked face sometimes doesn't match the lighting on the body or background. Look for shadows that fall in the wrong direction or skin tones that shift unnaturally.

Edge artifacts around the face. Look at the boundary between the face and the hair, ears, and neck. Blurring, flickering, or visible seams at these boundaries can indicate a face swap.

Mouth movement mismatches. When someone speaks in a deepfake video, the mouth movements may not perfectly sync with the audio, particularly on certain consonant sounds. This is subtle but detectable if you're looking for it.

Inconsistent background. The background behind a deepfaked person may shift, warp, or flicker slightly, especially during head movements.

Audio Deepfakes: What to Listen For

Unnatural pauses and pacing. Cloned voices sometimes have slightly irregular rhythm — pauses that are too uniform, breathing that doesn't sound right, or pacing that feels mechanical.

Emotional flatness. Current voice cloning technology is good at replicating the timbre of a voice but sometimes struggles with natural emotional variation. A voice that sounds technically correct but emotionally flat may be synthetic.

Audio quality mismatch. If someone claims to be calling from their phone but the audio quality sounds like a recording studio, or vice versa, something may be off.

AI-Generated Images: What to Look For

Hands and fingers. AI-generated images have historically struggled with hands — wrong number of fingers, impossible joint positions, strange proportions. This has improved significantly, but hands remain a weak point.

Text and lettering. Text in AI-generated images is often nonsensical or contains garbled characters. Look at signs, logos, labels, or any text visible in the image.

Symmetry and consistency. Check earrings (do they match?), clothing patterns (do they continue logically?), teeth (do they look natural?), and hair (does it behave realistically?).

Background inconsistencies. Objects in the background may merge into each other, have impossible geometry, or repeat in strange patterns.

Verification Strategies

Beyond looking for artifacts, develop verification habits.

Verify through a separate channel. If you receive a video call from someone asking for money or sensitive information, hang up and call them back on a number you know is real. If "your boss" emails you a deepfake video with instructions, walk to their office or call their direct line.

Check for original sources. If a viral video shows a public figure saying something shocking, check whether reputable news organizations are reporting it. Look for the original source. Check the date and context.

Use reverse image search. For suspicious images, use Google Reverse Image Search or TinEye to see if the image has been altered from an original.

Use deepfake detection tools. While not perfect, tools are emerging that can analyze media for signs of AI generation. Microsoft's Video Authenticator analyzes videos for subtle manipulation artifacts. Various browser extensions are being developed for image verification.

Building a Deepfake-Resistant Mindset

The most important defense against deepfakes isn't a technical tool — it's a shift in mindset.

We're entering an era where seeing is no longer believing. Video is no longer proof. A voice on the phone is no longer confirmation of identity. This is uncomfortable, but accepting it is the first step toward protecting yourself.

Adopt a verification-first approach to any unexpected communication that involves money, sensitive information, or unusual requests — regardless of how convincing the voice or video appears. Establish authentication protocols with family members and close contacts: a code word or phrase that confirms identity in situations where deepfakes might be used.

And be very careful about what you share online. Every photo and video of yourself that's publicly available is training data for someone who might want to create a deepfake of you. This doesn't mean you should never post anything, but it does mean you should be thoughtful about the volume and resolution of media you make publicly accessible.

Protecting Your Own Likeness

Beyond detecting deepfakes others create, you should also think about reducing the raw material available for creating deepfakes of you.

Limit high-resolution public photos and videos. Every clear photo of your face that's publicly available is a potential training input for a deepfake model. You don't need to disappear from the internet, but consider whether every photo you share needs to be public and in full resolution.

Be cautious with voice samples. Voice cloning requires as little as three seconds of audio in some modern tools. Voicemail greetings, public speaking recordings, podcast appearances, and even YouTube comments read aloud can provide enough material. If you're in a position where voice-cloning attacks could target you, consider shortening your voicemail greeting and removing unnecessary voice content from public platforms.

Establish family verification protocols. Agree on a code word or question with your family members that you can use to verify identity in suspicious situations. "If someone calls claiming to be me and asking for money, ask them what our code word is." This low-tech solution is remarkably effective against high-tech deepfake voice scams.

Watermark your content. If you share photos professionally, consider using invisible watermarking tools that embed tracking information in your images. This doesn't prevent deepfakes, but it can help establish provenance if your images are misused.

The technology will keep improving. Your awareness needs to keep improving too.