AI versions of me are popping up all over YouTube.
Not like fun little parodies or a memes. Actual deepfake avatars of me, hosting fake weather channels on YouTube, pumping out daily videos with my face, my voice, and completely made-up information layered over weather maps that don't even make sense. And they are getting views. Lots of them.
I started noticing something was off when my inbox blew up. People were emailing me asking if I had a second channel. Others were writing in asking "is this video AI?" about content I had never made. One person's granddaughter came running to them worried because she saw "me" saying something alarming about a storm that wasn't real. That one hit different.
The worst part? The videos are convincing enough that regular people can't tell. My own audience is getting fooled. And the content is pure garbage. It's AI slop, nonsense narration over unreadable maps, but because it looks and sounds like me, people trust it. That's the whole game. They're borrowing the trust I've spent years building and using it to farm clicks with junk content.
I reported the channels. My YouTube rep is on it. We're flagging everything we can. But here's the reality: you take one down and another pops up. The tools to make these videos are getting cheaper, faster, and better every single day. It's a game of whack-a-mole and the moles are multiplying.
So I did what any reasonable person would do. I built a fruit-based verification system.
Yeah. You read that right.
How It Works
Every morning at 8:05 AM, before I even think about making a video, a script runs on my server and randomly picks one of four items: apple, banana, grape, or pickle. (Yes, pickle. It's a fruit.)
That selection gets locked in and stored. Nobody sees it yet, not even me until I check. Then when I go to film my video, I grab whatever fruit got picked that day, and I take a bite of it right there on camera at the very beginning.
If you ever want to verify that you're watching the real me, you go to ryanhallyall.com/real. The site shows you what today's fruit is. If the fruit I'm biting in the video matches what's on the site, you're watching a real Ryan Hall video. If it doesn't match, or if there's no fruit bite at all, it's fake.
The fruit doesn't get revealed on the site until after I upload the video, so there's no way for a deepfake channel to cheat by checking ahead of time. The whole thing is time-locked and automated.
I Know How This Sounds
Trust me, I am fully aware that I just described an elaborate tech stack involving multiple servers, automated pipelines, and a React frontend with a spinning lock animation... all so that people can confirm whether or not I bit a grape today.
The tech behind it is honestly kind of overkill for what it does. We've got Python scripts, Cloudflare storage buckets, a system that polls the YouTube API waiting for my uploads, timezone-aware date matching, a state machine tracking whether the fruit has been revealed yet. It's a lot of infrastructure dedicated to produce.
But that's kind of the point. The fact that I had to build something this absurd just to prove my videos are real should tell you everything you need to know about where we are right now with AI and deepfakes.
This Is Not the Solution
I want to be super clear about something. This is not a real solution. It's a band-aid. It's a creative, slightly ridiculous band-aid, but it's a band-aid.
Right now, if you're a creator dealing with deepfakes of yourself, your options are basically: report the channels and hope the platform acts fast, use whatever detection tools exist (YouTube just rolled out a likeness detection feature, which is a start), and get creative. That's what this is. Me getting creative.
But the deepfakes are going to keep getting better. I said it on Twitter and I'll say it here too: my prediction is that eventually the fake videos will be better than my real ones. That's not even a joke, that's just where the technology is headed. At some point, biting a piece of fruit isn't going to cut it because the AI will be able to generate a realistic video of me biting a piece of fruit too.
We need actual infrastructure-level solutions for this. Content authenticity standards, verification built into platforms, maybe something with digital signatures or provenance tracking baked into the video files themselves. I don't know exactly what the answer looks like, but I know it's not one guy chomping on an apple in his studio.
Why I'm Doing It Anyway
Because right now, today, it works. It's simple enough for anyone to understand. You don't need to be tech-savvy. You don't need to analyze pixels or run detection software. You just look at the fruit.
If me standing here biting into a pickle at the start of every video gets even a few more people thinking about the deepfake problem, talking about it, being more skeptical of what they see online, then it's worth looking a little silly.
The conversation matters more than the fruit. The fruit is just how I'm starting it.
What You Can Do
If you see a video of "me" that doesn't have the fruit bite, don't trust it automatically. Check the channel name. My real channel is Ryan Hall, Y'all. That's it. There's no second channel, no "Severe Storm Center" spinoff, nothing like that.
If you spot a fake channel using my face, report it. Flag it on YouTube. Send it my way. We're keeping a running list and working with YouTube to get them taken down as fast as possible.
And look, this doesn't just affect me. If they can do it to a weather guy, they can do it to anyone. Your local news anchor, your favorite teacher on YouTube, your uncle who makes fishing videos. The tools are out there and they're only getting easier to use. Stay skeptical. Verify what you watch. And if someone isn't biting into a random piece of fruit at the start of their video... well, at least now you know why I am.
Check today's verification at [ryanhallyall.com/real](https://ryanhallyall.com/real)
