Your Eyes Are Lying to You: A Field Guide to Spotting AI-Generated Fakes Online

Let's be honest with each other for a second. You've probably shared something online that turned out to be fake. Maybe it was a doctored photo. Maybe it was a video that seemed just a little too perfectly timed. If that's happened to you, you're not alone — and you're not dumb. The technology behind synthetic media has gotten frighteningly good, and our brains simply weren't built for this.

But here's the thing: you can get better at this. Not perfect — nobody is — but meaningfully better. And that matters a lot, because the people creating convincing fakes are counting on you to scroll fast and share fast.

At SkeptiCamp, we believe the best defense against manipulation isn't some fancy software. It's a trained, skeptical mind. Let's build yours up.

What Exactly Are We Talking About?

The term "deepfake" gets thrown around a lot, but it covers a pretty wide range of stuff. At the simpler end, you've got AI-generated still images — faces of people who don't exist, fake screenshots of news headlines, or real people placed in scenes they were never in. At the more complex end, you've got video deepfakes: footage where someone's face has been swapped onto another person's body, or where a real person appears to say things they never actually said.

There's also a growing category of AI-generated audio — voice cloning tools that can replicate someone's speech patterns with just a few minutes of sample audio. In 2024, robocalls using a cloned version of President Biden's voice circulated before the New Hampshire primary, urging Democrats not to vote. That's not science fiction. That already happened.

Synthetic media is no longer a niche problem. It's a mainstream one.

The Red Flags Your Brain Tends to Miss

Human perception is surprisingly bad at catching certain kinds of visual inconsistencies, especially when we're moving quickly through a feed. Here's what to slow down and look for:

On still images:

• Weird ears and hair edges. AI image generators still struggle with fine detail at the boundaries of faces. Look closely at where hair meets skin, or how earrings sit. If something looks slightly melted or blurred, that's a flag.
• Background incoherence. Text in the background of AI images is often garbled or nonsensical. Bookshelves might have spines with unreadable titles. Signs might have letters that don't form real words.
• Hands and fingers. This is the classic tell. AI-generated hands frequently have the wrong number of fingers, fused digits, or fingers bending at impossible angles. Always check the hands.
• Lighting that doesn't match. If a person's face is lit from one direction but the background suggests light from another, something's off.
• Skin texture that's too smooth. Real human skin has pores, asymmetry, and subtle variation. Heavily AI-generated faces often look like they've been airbrushed to an inhuman degree.

On video:

• Unnatural blinking. Early deepfake models didn't blink at all. Newer ones do, but the rhythm can still feel off — either too infrequent or weirdly mechanical.
• Facial edges that shimmer or blur. Watch the border between a person's face and their hair or neck during movement. Deepfakes often show a subtle flickering or smearing effect there.
• Mouth sync problems. When someone speaks, their lips, jaw, and tongue all move together in complex ways. Deepfake audio-video sync frequently breaks down on certain consonant sounds.
• Emotional mismatch. The voice might express anger, but the facial muscles don't quite follow. Genuine emotional expression is incredibly nuanced, and fakes often flatten it.

Free Tools You Can Actually Use Right Now

Your eyes and brain are your first line of defense, but you don't have to go it alone. Several free verification tools are available to anyone with a browser:

• Google Reverse Image Search / TinEye — Drag any suspicious image into these tools to see where else it appears online. A "breaking news" photo that shows up on a gaming forum from three years ago is probably not what it claims to be.
• FotoForensics (fotoforensics.com) — This free tool runs Error Level Analysis on JPEGs, which can reveal areas of an image that have been digitally altered. It takes a little practice to interpret, but it's a genuinely useful skill to develop.
• Hugging Face Deepfake Detectors — The AI research community has built and published several open-source deepfake detection models. Some are available through the Hugging Face platform without any coding knowledge.
• InVID / WeVerify Plugin — Originally designed for journalists, this browser extension lets you break videos into frames for reverse image searching, check metadata, and run other verification checks. It's free and works on Chrome and Firefox.
• Sensity AI — Offers limited free deepfake detection for video files. Used by newsrooms and researchers.

None of these tools are foolproof. A sophisticated fake might pass all of them. But most fakes you encounter in the wild aren't sophisticated — they're quick jobs built to fool people who aren't looking closely.

The Critical Thinking Layer That Tools Can't Replace

Here's where the SkeptiCamp philosophy really kicks in. Tools help, but the habits of mind matter more.

Ask who benefits. Before you share something outrageous, ask yourself: who gains if people believe this? Fake content is almost always created with a purpose — to sow confusion, to damage a reputation, to sell something, or to push a political narrative. Identifying the potential motive doesn't prove something is fake, but it's a useful starting question.

Check the source chain. Where did this video or image actually originate? If you can only trace it back to an anonymous account that was created last month and has no other posts, that's a significant red flag. Legitimate newsworthy content almost always has a traceable origin.

Slow down before sharing. Research consistently shows that the simple act of pausing — even for a few seconds — before sharing reduces the spread of misinformation. Your instinct to share something shocking is exactly the instinct that bad actors are trying to trigger. Resist it.

Seek corroboration. If a video shows something genuinely significant, major news outlets will be covering it. If you can't find a single reputable source reporting on the same event, that's meaningful.

Accept that you can be fooled. This one's uncomfortable, but it's essential. Overconfidence in your own ability to spot fakes actually makes you more vulnerable, not less. Staying a little humble keeps you alert.

This Is a Skill, Not a Superpower

Spotting synthetic media isn't something you either have or don't have. It's a skill that develops with practice, like anything else. The more you consciously apply these checks, the more automatic they become.

We'd encourage you to bring this into your next SkeptiCamp workshop or community session. Practice running images through FotoForensics together. Watch a deepfake video as a group and try to identify the tells before revealing the answer. Make it a game. That kind of active, social learning is exactly how critical thinking skills actually stick.

The technology behind synthetic media is going to keep improving. The fakes are going to get harder to catch. But so can you — if you keep practicing.

Don't trust your first instinct. Verify. Question. Then share.