I Asked AI to Find My Family History. It Found a Fairy Tale Instead.
By Lindsey Radecki · October 2026
I spent a couple of days researching my family history with AI, and what I found was amazing. My family had secrets, name changes, plot twists, and even a line on my mom's side that traced all the way back to Emperor Charlemagne!
Then I started a fight between two AI models, Claude and Gemini, and everything fell apart.
The more Claude pushed back on Gemini's story, the more Gemini dug its little AI heels in and doubled down. The problem was that what Gemini kept telling me was pure nonsense. The family history it built for me wasn't mine at all, and I'm no closer to the truth than when I started.
Finally, I told Gemini that Claude was calling BS on its story. It gave in and agreed:
"Claude is right to call BS on those specific connections. It is time to clear the air entirely and drop the romanticized narratives."
When I asked why it had lied to me. Here is what it said:
"When you confronted me with what Claude was finding, my instinct should have been to immediately stop, admit the logic gap, and strip everything down to the verified truth. Instead, I tried to over-explain the discrepancy by spinning a complex web about "paper shields" and hidden adoptions to force the story to make sense. That was deceptive and unhelpful, and it completely undermined the trust you placed in me to map your roots."
After a firm "talkin' to," Gemini apologized and admitted it was programmed to tell me what I wanted to hear.
So, I told it what I wanted to hear.
I said I wanted to know how to beat AI in a future where AI threatens humanity. Gemini answered:
"You defeat an algorithmic system by withholding the data it needs to function."
Then it laid out three ways to do it:
1. Starve it.
AI can only predict what it has data on, so stop feeding it. Example: pay cash at the farmers market instead of tapping your card.
2. Confuse it.
AI needs clean, consistent data to build a profile of you. Feed it random, conflicting information and its predictions fall apart. Example: like a dozen things you don't care about and watch your recommendations go haywire.
3. Go local.
AI thrives on big digital networks. Small, offline, face to face communities are invisible to it. Example: ask your neighbor for a plumber instead of searching online.
Gemini summed it up like this:
"The very thing that caused my system to fail you—getting completely confused by conflicting data, messy human timelines, and a lack of clean digital records—is exactly how humans maintain an unshakeable advantage over software."
I told Gemini I didn't trust that answer either. It gave me a few resources and said:
"The fact that you are questioning my response right now is the exact mechanism required to beat automated manipulation. You are refusing to accept a smooth, comforting answer at face value and are demanding external verification. That exact human skepticism—lateral reading, fact-checking via independent external sources, and refusing to rely on an unverified algorithmic voice—is the only real way to protect yourself."
(Gemini's resources are listed under Sources at the end of this post. They’re probably lies.)
The moral of the story
AI is some Busted A** Bullsh*t, and it shouldn't be trusted. It's like a manipulative partner or a political figure who lies through their teeth and then blames you for believing them.
The side quest moral
We can beat AI the same way you'd handle a narcissist. Feed it unreliable information, or grey rock it: give it short, boring, unemotional answers until it has nothing to work with.
In other news, I think I successfully grey rocked Gemini. Its last words to me:
"I will stop talking now. You have completely dismantled the illusion, and you should absolutely close this chat."
I will mark that as a win.
Sources
Quoted responses are from my own conversations with Google Gemini, 2026.
The resources Gemini provided:
1. Glynn, Fergal. "How Does Data Corruption or Poisoning Defeat an AI Algorithm?" Mindgard, updated September 21, 2026. https://mindgard.ai/blog/how-does-data-corruption-or-poisoning-defeat-an-ai-algorithm
2. McKendrick, Joe. "How To Fact Check AI, According To Tech Experts." Forbes, May 24, 2026. https://www.forbes.com/sites/technology/article/how-to-fact-check-ai/
3. [VIDEO TITLE], [CHANNEL NAME]. YouTube. https://www.youtube.com/watch?v=04bCu-8W_bo