How to Spot a Deepfake: 1 School Photo Is All It Takes

Here's the sentence that should stop you mid-scroll: a girl's yearbook photo, the same one printed in 400 copies and handed out at her middle school, got run through a free app and turned into a fake nude image before her mom even knew the app existed. No hacking. No stalker with special skills. Just a normal photo, a free app, and a few taps. That's the whole story now, and knowing how to spot a deepfake — an AI-made fake photo that looks real but isn't — is quickly becoming something every parent needs, not a niche tech skill.

Forget the idea that this is a Hollywood problem, something that happens to famous people with a million photos online. It's not. It's a classroom problem now. CBS News reported that teens are facing a real surge in AI-altered images as these apps spread across app stores and group chats, and the tools doing the damage aren't buried on some sketchy corner of the internet. They're on the same app store as your kid's homework planner.

The One Photo Problem Behind AI Altered Teen Images

Let's talk numbers, because they tell the story better than I can. Last year there were more than 21,000 fake pornographic videos made with AI floating around online — a jump of over 460% compared to the year before. This year, more than 143,000 new ones showed up, which is more than every previous year combined, put together. That's not a trend. That's a landslide.

And here's the detail that should really land: UNICEF found that at least 1.2 million children around the world said, when asked directly, that their own photos had been turned into sexual images using AI in just the past year. Not photos of strangers. Their own faces. Their own bodies, faked.

Sit with that for a second. One in 17. That's not a statistic from a far-off future. That's a real number of kids sitting in a real classroom, right now, today. CBS News actually went and tested this instead of just writing about it — they ran 70 different AI photo-editing apps through their paces and found that 19 of them turned ordinary photos into disturbing, sexual content, including nudity. Many of those apps have zero age checks. A 13-year-old can download one just as easily as a 30-year-old.

It's Not Always a Stranger

Here's where it gets uncomfortable. Research from Common Sense Media found that 1 in 5 teens admits to having personally made an AI-generated sexual image of someone else, and 1 in 4 has either shared one or knows a classmate who has. Read that again — this isn't just a "predator lurking online" story. Sometimes it's the kid two seats over in algebra. The tools are that easy to use, and the social consequences, for now, are that thin.

Right now, the cleanup falls to the person with the least power to manage it: the kid in the image. — reporting by NBC News

That line stuck with me because it's just true. A teenager can't take down a group chat. A teenager can't file a subpoena. A teenager can barely get an adult in the building to believe them fast enough to matter, and by the time anyone does, the image has already made three more laps around the school's group chats.

How To Spot A Deepfake Before Any Damage Spreads

I'm not going to pretend there's a magic trick that catches every fake — the apps get better every few months, and the tells that worked last year (weird hands, blurry teeth, mismatched lighting) are already fading out. But a few things still hold up. Look at the edges around hair and ears — AI still struggles to blend those cleanly. Look for skin that's too smooth or too shiny in patches, like a filter got applied unevenly. And honestly, the biggest tell right now isn't visual at all — it's context. Ask yourself: does this image match where this person actually was, what they actually wore, who they were actually with? Fakes often show up detached from any real event, floating with no believable backstory. Continue reading: How To Spot A Deepfake 1 School Photo Is All It Takes.

If you've ever felt that little itch of doubt looking at a photo of your kid's friend, or a screenshot someone forwarded you, or a "did you see this?" text — trust that itch. That's exactly the question all of this AI-authentication technology exists to answer: is this real, or was it made? One practical thing you can actually do tonight, before any app or service gets involved: check whether your child's school posts team rosters, yearbook pages, or class photo directories publicly online, fully indexed and searchable by anyone. Most schools default to "public" because nobody asked them to change it. Ask them to. Fewer public, easily-scraped photos means less raw material for someone to grab in the first place.

What Schools Still Get Wrong About AI Altered Teen Images

Look, nobody's saying this is simple to fix. But most schools are still treating this like a cyberbullying problem from 2015 — write a policy about "inappropriate content," send a letter home, move on. That doesn't cut it anymore, because the harm here isn't just embarrassment. It's a fake image that looks convincingly real, showing a body that isn't actually theirs, seen by classmates who may not stop to question whether it's fake at all.

The federal Take It Down Act was signed into law on May 19, 2025, and took effect on May 19, 2026, requiring platforms to remove this kind of content once notified. That's a real step. But here's the catch — laws punish what already happened. They don't stop a 14-year-old from opening an app store, typing "photo editor," and finding one of those 19 apps that CBS flagged. Legislation lives downstream of the harm. The apps live upstream, right where the photo still exists.

So here's my actual question for you tonight, and I want you to sit with it instead of scrolling past: if your kid's school found out tomorrow that a fake image of a student had been made and shared, would anyone there know what to do first — who to call, what to preserve, who to protect? Because right now, in most districts, the honest answer is no. And a policy written after the fact doesn't un-send a photo that's already been forwarded 40 times before first period.