Are AI Doctors Easy To Generate? (beta β)

It's actually very easy to AI generate a realistic doctor. I'm not talking about the silly obviously AI accounts or some fake persona with a madeup name. I'm talking AI generating real doctors. So, someone with a real medical degree, a real career, real patients, and having AI use their face and voice to say things the real doctor would not say.

But before I show you how easy it is, I want to get something straight. I am against deep faking real doctors without their consent. And I strongly discourage deep faking any doctor, real or made up, without a real human doctor accountable for what they say. Human accountability is extremely important because humans fear the consequences of what they say.

So they double check. AI does not the lack of AI accountability is a huge problem for giving out what is essentially medical advice, even if there is a disclaimer. Now, while that sounds like a problem in theory, we need to answer the question, does AI actually hallucinate bad medical advice? I've covered this in previous videos and I'll continue to cover in future videos, but yes, even the most advanced chat bots currently under triage emergencies, possibly leading people to their deaths.

For example, a research paper from Mount Sinai in May 2026 found that Chad GPT under triage 52% of emergency cases. That means that people were told it's okay when it wasn't. I also found similarly dangerous medical misinformation by a channel I suspect is AI, but that's another video. All that is to say, unreed medical information is dangerous, especially to those who are unaware that it's produced by AI.

So, you probably think you'd spot an AI doctor and just tune it out. You're probably wrong. What you've been watching for the past minute is AI. I'm not real. This is a selfdeep fake. It's my face, my voice, my mannerisms, but none of this is me sitting at a camera. Every frame you've seen so far was generated.

Nas reviewed the script, but he just as easily could have skipped it. And this didn't take much work. I'll tell you more about that in this video, but first, let me say it could be even better if Nas paid more for it. Here's what I mean. Wild, right? So, Nas ran an informal test with friends and family, people who know what he looks like.

About 60% couldn't tell. And these are people who had a reason to look closely. And for the 40% who did catch it, most of them said it was the eyes. So to stay transparent and accountable, I'm tagging the rest of this video as AI generated because that's the difference I disclose. The channels you should be worried about don't.

If 60% of people who actually know NAS couldn't spot that it was AI generated, just imagine how high that percentage gets when it's non-technical people looking at a doctor they've never even seen in real life. And that's already the case with a lot of doctors right now. And it's only going to get worse.

Nas doesn't want to make it worse, so this video isn't going to be a tutorial. The goal is to help you recognize AI generated doctors. There's a difference between knowing a lock can be picked and knowing how to pick it yourself. Nas just wants you to know that the lock can be picked. NAS builds AI software for a living in healthcare.

He works with the real systems, the real limitations, and the real ethical lines that get crossed when this kind of tech enters a clinical setting. NASA's been studying AI doctor channels for months now. He knows what the good fakes look like. He knows what they don't look like and he's made me narrate this video while he shows you how it was built.

So that's why Nas is showing you this, not so you can build one, but so you can spot one when it's pointed at you. And he's building the whole thing on himself for an ethical reason I'll keep saying out loud. Starting with the very first move, who do you deep fake? The operations that do this at scale don't pick random people.

They pick real doctors who already have some kind of online presence. a LinkedIn page, a published paper, a hospital bio, a podcast appearance, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel, an appearance on a TV channel. So, that was a lot of those.

It could be anything. Why? Because when someone searches the name and finds the real doctor, they just assume the channel is them. So the real person does all the trust work and the operation collects the reward. Now Nas could do that. He has the technical ability, but impersonating a real doctor without their consent is illegal.

It's not just a terms of service violation. So in the United States, impersonating a real doctor can count as identity fraud, defamation, and in states with right of publicity laws, civil liability that follows the operator, even through a fake LLC. In the UK, it's fraud by false representation under fraud act 2006.

But Nas deep faking himself is a completely different thing. There's no real person being impersonated, so there's no victim. And as far as Nas could find, no law it actually breaks. That's the whole reason the entire build is on him. So Nas is deep faking himself. Not because it's the easy path, but because it's the only defensible one.

And once you've picked who to fake, the next job is making them feel like a real person. So every convincing defig runs on what's called a character bible. It's basically a document that describes who this person is. physical description, specialty, medical school, where they trained, their communication style, their preferred words, their verbal ticks, family members they reference, the city they live in, the patients they care about, and the detail isn't decoration, it's consistency.

A deep fake doctor who says different things about their background across videos gets caught. But a deep fake doctor whose every answer comes from that same character document sounds real. So the persona is what makes the fake feel like a person. And they're also used in a very cunning way on the scripts.

So the scripts that get people are built on patient stories, personal anecdotes. A patient came to me with this exact problem. Here's what I told them. And those stories are always vague. No name, no date, no identifying details. But the real reason isn't privacy. It's that there was never a patient.

You can't give specifics about someone who doesn't exist. And here's why. You can't just call it out. In the US, there's a privacy law called HIPPA, and most countries have their own version, and it means a real doctor would have to strip those same details out anyway. So, a completely invented story and a real one with the privacy scrubbed off look exactly the same.

You can't check if it happened. You can't find the patient. You can't ask the hospital. And the second you push, the channel just says it's protecting patient privacy. So, they're not abusing HIPPA. they're weaponizing it. A rule built to protect real patients is the perfect cover for inventing fake ones.

I want you to remember that pattern. So, personal medical stories are perfect for AI factories because they never have to prove a single one. But a script is just words on a page until something says them out loud. And that's where the voice comes in. So, voice cloning is the part that surprises people the most.

And not because it sounds good, it's because of how little it actually takes. So Cartisia clones a voice from 3 seconds of audio. 11 Labs needs about 10. And Speechify needs no account at all, just a quick recording in your browser. Nas used his own voice. That's the ethical reason again. And for anyone targeting a real doctor, they've got podcasts, YouTube videos, conference talks, medical school graduation speeches.

Most doctors have their voices in conference archives, podcasts, professional proceedings, more places than they could remember. So, a convincing voice takes a few seconds of audio and an afternoon to set up the documented time on every platform. N checked. So, the voice takes that little You think the face is the part that finally takes real time and real money.

It mostly doesn't. So, I'm just going to show you what Nause got in the order he actually did it. each attempt better than the last. It all starts with a picture. The first time Nas just used a regular photo of himself, the kind anyone already has, sitting in front of the camera. The tool fills in the rest in about 2 minutes, and there's no question it's him.

This is the video he got out of it. The next tactic is the part where you stay until the end of every video, even when you don't have time. Second attempt. You don't even need a real photo. You can generate one. So Nas generated a picture of him with scrubs and then turned that picture into a video where you stay until the end of every video even when you don't have time.

The point is you can make a convincing doctor out of almost anything. And if you pay a bit more, it can look even more realistic. NAS generated one more picture in this high quality and had the AI do a bit more work on the video. And you're seeing that version right now. So, you can make really amazing quality videos with very little effort.

If you think it takes a lot of AI skill, well, I've got news for you. But real quick, because this is the part that actually protects you. Everything Nas has learned about telling a real doctor from a fake one, he put in one place, the AI health trust vault. If you watch any health content online, that's the link to grab from the description.

Okay, back to the news. Those are all real. Most of them have free tiers. Several work from a single photo and some work in real time on a live call. Now, I'm not recommending all of them. Honestly, several are genuinely bad, but they all exist. They're all available. And together, they make one thing clear.

You don't need technical skill to do this. You need an afternoon. So, that's the part that should bother you. Not one tool, dozens. The tools were never the barrier. The barrier was just knowing it was possible. And those same tools throw in the final touch for free, dropping the fake doctor into any room you want.

So, a convincing doctor doesn't stand in front of a gray wall. They stand in front of a hospital or a home office or a consultation room. And most of the avatar stacks NAS looked at let you swap the background in a single click. No green screen, no re-recording. Same face, same voice, different room, different context.

So, a few backgrounds and a deep fake doctor suddenly has a life. So, a few backgrounds and fake doctor suddenly has a life. And here's the thing I want you to sit with. These tools are going to keep getting better. So, with a couple backgrounds, the fake doctor has a whole life in any room you want.

And the few tells still left today in the eyes, the hair, the hands, those are closing, too. So, let's add up everything it actually takes. So, three attempts to something 60% of people who know NAS couldn't identify as AI. A persona document that keeps every answer consistent. vague patient stories that are legally impossible to disprove.

A voice that needs 3 seconds of real audio to clone. A face that needs 15 seconds of real video. And dozens of platforms that make all of it accessible for the price of a streaming subscription. So, the system isn't hard to build. It isn't expensive to run and is already running on channels with hundreds of thousands of people watching.

And what it looks like when you have factory money is even better than what you've been watching for this video. A doctor talking about health calmly, confidently with personal stories you can't check, credentials you can't verify, and advice that might actually hurt you. So, [snorts] the deep fake itself isn't the danger.

It's the proof that the verification layer, the thing that used to let you trust a face, is gone. When the face, the voice, the credential, and the patient story can all be manufactured in a week for the cost of a gym membership, the question isn't whether it's real anymore. The question is how you tell.

And there are still a few things these tools can't do. And those are exactly how you catch them. I'm going to go through exactly what those are in the next video. The tells that survive every iteration. The ones that are closing and the ones that aren't. If the question this video raised in your head is, "How do I know if what I'm watching is real?"

That next video is the