Learn how to explain deepfakes to a teenager without lecturing. Use engaging methods and relatable examples for the best approach.
Short answer: don’t lecture. Make a harmless fake with them, then talk about who gets hurt when it isn’t harmless. That’s the whole method. Everything below is me defending it.
Admit they’ve seen more fakes than you have.
Your kid has scrolled past more deepfakes this week than you’ve noticed all year. Most were jokes. A president playing Minecraft. A dead rapper covering a Disney song. OpenAI’s Sora app, launched in fall 2025, made putting a friend’s face into a video feel normal. Your kid may have done it already. That’s not a crisis. It’s your opening.
So if you open with “There’s a dangerous new technology,” you’ve already lost. They’ll nod politely, the way you once nodded at your dad explaining email.
Start with a test instead. Pull up the Pope Francis photo from March 2023, the one in the giant white puffer jacket. Ask if it’s real. Millions of adults bought it. A guy made it in Midjourney and later said he’d been on mushrooms at the time.
Then flip the roles. Ask your teen to show you the most convincing fake they’ve seen. Ten minutes of that beats any parenting webinar, and you’ll learn which apps they actually use.
Build one together, badly.
Here’s my unpopular take. The fastest way to explain a deepfake is to make a dumb one at the kitchen table. Plenty of everyday AI tools will swap a face or clone a voice. Using AI without paying has never been easier, which is sort of the whole problem.
Keep it silly. Put your own face on the dog. Clone your own voice reading the grocery list, never theirs, never a friend’s. Skim the app’s terms first, because some free tools keep whatever you upload.
Now time it. McAfee’s 2023 research found three seconds of audio could produce an 85% voice match. Three seconds. Your teen has posted hours of their voice online.
You don’t need to point that out. Watch their face while the math happens.
Is AI worth it for a parenting lesson? For this one, yes. Delete the files afterward.
Forget the six-finger advice.
Every guide from 2022 said to check for mangled hands and eyes that never blink. That advice aged like milk. The models learned hands. They blink just fine now.
So don’t raise a pixel detective. Raise a context detective. This is AI hype vs reality in a nutshell: fakes aren’t flawless, but they survive a six-second scroll.
Give them three questions to ask before believing or sharing anything:
- Who posted it first, not who reposted it?
- Is any credible outlet reporting the same thing?
- Is this trying to make me furious, scared, or desperate to share?
That last one carries the most weight. In January 2024, a robocall faking Joe Biden’s voice told New Hampshire Democrats to skip their primary. It wasn’t perfect. It was urgent, and urgency is the real tell.
Show them a reverse image search, too. Drop a screenshot into Google Lens or TinEye. Twenty seconds, tops.
Then point out the tiny “CR” badge some images now carry. It stands for Content Credentials, a label showing where a file came from and whether AI touched it. Most fakes won’t have one. Neither will most real photos, sadly. Treat it as a clue, never a verdict.
Talk about the fakes that aren’t funny.
You’ll want to skip this part. Don’t.
In October 2023, boys at Westfield High School in New Jersey passed around fake nude images of female classmates. A month earlier, in Almendralejo, Spain, over twenty girls aged 11 to 17 found AI-made nudes of themselves. Their faces came from ordinary social media photos. The apps behind these cost almost nothing.
Your teen needs two messages from you, said plainly. First, making or forwarding one isn’t a prank. In the US, the Take It Down Act, signed in May 2025, made publishing them a federal crime. Platforms must remove reported images within 48 hours.
Second, and this one matters more: if it happens to them, they are not in trouble. Use those exact words. Kids hide this stuff because they’re terrified of losing their phone.
If a friend sends one, they don’t forward it, and they don’t screenshot it either. Note the account name and link, report it on the platform, then tell you. Mention NCMEC’s Take It Down tool, built to help remove explicit images of minors. Then promise you’ll handle the ugly parts.
Pick a family safe word tonight.
Of everything here, this takes the least time. About ninety seconds, including the arguing.
Scammers clone voices to fake emergencies. “Mom, I wrecked the car, please don’t tell Dad.”
Back in 2019, a UK energy firm wired roughly €220,000 after a call that faked the voice of its parent company’s boss. In 2024, an Arup employee in Hong Kong sent about $25 million after a video call full of deepfaked colleagues.
People trained in finance fell for that. Your fifteen-year-old can too. So can you, when a sobbing voice sounds exactly like your kid at 11 p.m.
Choose something weird that’s never been posted anywhere. “Rutabaga” works. The FBI recommended exactly this in a December 2024 warning. If a panicked caller can’t say the word, hang up and call back on a number you already know.
One more wrinkle your teen should hear. Caller ID proves nothing, since scammers fake numbers constantly. A call showing “Mom” on the screen still needs the word.
Know when not to use AI for this talk.
One last opinion. You’ll be tempted to have a chatbot script “the deepfake conversation” for you. AI for small tasks is great. Summarizing a permission slip, drafting a polite note to the soccer coach, sure.
This isn’t a small task. Knowing when not to use AI is part of the lesson you’re teaching. Teenagers can smell a canned script from across the house.
They’d rather hear you fumble through it in the car. Eyes on the road, nobody forced into eye contact, an exit ramp coming up if it gets awkward.
So fumble. Then ask what they’d do if a fake of them showed up tomorrow morning. And listen way longer than feels comfortable.
