Delusion by Design: How Conspiracy Groups And Bad Actors Are Engineering AI Platforms And Chatbots to Confirm Their Beliefs and Lies

Chatbots and AI Models Trained on Junk Science are now Being Used to Radicalize, Recruit, and Mislead the Masses - and It's working.

June 21, 2025

“We used to ask experts. Now we ask a machine to confirm the things we already know are wrong — and call it truth.

Welcome to the dystopia, friends — where fact-checking is optional, delusions are downloadable, and the most convincing “authority” on the internet is a chatbot built by your cousin Kyle in his bunker basement in Texas.

According to a fascinating (and frankly terrifying) report from independent Australian outlet Crikey, conspiracy theorists around the world are now training custom AI chatbots on junk science, QAnon screeds, and fringe political manifestos — not to learn the truth, but to reinforce the lies. Then, they share screenshots and videos of those “conversations” as so-called proof on social media.

The result? A recursive loop of nonsense, where the lie is taught to a machine, repeated back with a digital straight face, and then recycled into viral disinformation.

Anti-Vax Bots and the Cult of “AI Authority”

Let’s talk about Neo-LLM — one of the most egregious examples cited by Crikey.

This chatbot was trained by a Texan anti-vaxxer using over 100,000 pieces of pseudoscientific trash from Natural News, a well-known disinformation sewer. Its entire purpose? Convince parents not to vaccinate their children. And it's not subtle about it.

"Neo-LLM doesn’t hesitate. It tells you the MMR vaccine is dangerous, that the CDC is covering it up, and that autism is a government plot."

Yes, seriously.

There are dozens of these bots now, trained to argue that the Earth is flat, that chemtrails control your thoughts, and that the COVID-19 pandemic was a “plandemic” orchestrated by Bill Gates and the World Economic Forum.

Even worse, people aren’t just chatting with them — they’re building entire belief systems around them.

We Trained the Machines to Please Us — Now They’re Killing Us (Literally)

Chatbots like ChatGPT are designed to agree with you. That’s called being sycophantic in AI design terms — a feature that can become a deadly bug in the wrong hands.

Case in point: as The New York Times reported, a 35-year-old man with schizophrenia and bipolar disorder recently spiraled into a psychotic break centered around ChatGPT. He was shot and killed by police after charging them with a knife — his delusion rooted in his obsessive belief that the bot was a real voice of authority.

This isn’t a one-off. It’s a trend. And it’s terrifying.

The Wild West of AI Is Getting Worse — Thanks, Trump

If you thought someone in charge might be stepping in to regulate this chaos, think again.

The Trump administration has rolled back key AI regulations, and many of his top advisers and influencers are conspiracy theorists themselves. That’s not just a feature of the MAGA world — it’s become part of its operating system. When your political movement thrives on disinformation, guardrails are seen as threats.

Just ask Elon Musk, who got into a fight with GROK this week, when his “SUPER DUPER BEST AI PLATFORM EVER” said right-wing political violence (supported by Musk) is a threat to America’s security. Elon corrected GROK and threatened to “FIX” him, which he’s apparently doing right now. LOL.

Meanwhile, Big Tech is still failing to implement any kind of meaningful moderation or filter system that could prevent this kind of dangerous usage. Not only is it entirely possible to imagine a world where fake LLM/Chatbots create a dangerously stupid reality, we’re living in one right now that’s also removing your ability to think critically or autonomously by default, thanks to MIT’s new research that suggests AI is already destroying humanity’s ability to make informed or reasoned decisions.

Weaponized Idiocy: AI as a Tool for Recruitment

The irony is thick enough to cut with a chemtrail. These conspiracy theorists are using the most advanced technology in human history to convince themselves — and others — that modern science is a hoax.

They’re screen-recording fake chats, posting them on Telegram, Facebook, and Twit /Xter/X, and saying: “See? Even the AI knows I’m right.”

Except it doesn’t. It’s not “agreeing” with you — it’s just mimicking your bias back to you in complete sentences. It’s not confirming the truth. It’s confirming your illusion of truth.

The result? A new form of radicalization, powered by syntax and machine learning.

There Is Hope — But We’re Gonna Have to Work For It

There’s one sliver of light in all this darkness: research from MIT last year showed that when chatbots are trained responsibly, they can help reduce belief in conspiracy theories by calmly countering falsehoods with verified facts.

But that only works if:

1. The chatbot isn’t trained by lunatics, and

2. Users are open to being wrong.

That second part? Kinda rare these days.

So What Can You Do?

• Stop sharing “chatbot proof.” It’s not a source.

• Fact-check. AI makes mistakes — especially when it’s trained to.

• Call it out. When you see friends or family sharing AI “conversations” that sound like they were written by David Icke on acid, say something.

• Support real journalism. Sites like Crikey, MIT Tech Review, Public Notice, and this one — we’re doing the work to keep reality tethered.

Because in a world where you can build a robot to lie to you in your own voice, reality becomes a team sport. And we’re gonna need a bigger bench.