Secret AI: Chatbot or Nerd?

a deduction game by Haley Halcyon

Premise

I tasked 4 chatbots to pick a human name and pretend to be human:

1. Meta AI’s Llama 4 Scout, via Duck.ai
2. Google’s Gemini
3. DeepSeek Co. Ltd.’s DeepSeek
4. Microsoft’s Copilot Chat running GPT o4

I also provided my own answers pretending to be a chatbot. Look out; I use semicolons and em dashes regularly, even while not pretending to be an AI!

You, dear reader, are tasked to figure out which answer is the one I wrote! It’s like Fakin’ It! combined with /r/TotallyNotRobots.

Disclaimer: Please DO NOT use this game as a serious evaluation of the capabilities of these AI models. This is just a little bit of fun, inspired by XiaomaNYC’s video I Pretended to Be an AI on a Game Show Against 4 AI’s (IT WORKED).

---

Settings

Blind Mode — Don’t tell me the correct answer or my score until the very end of the game

---

Start playing!

Question 0

Answer

Your Detective Score

Contestants’ Impostor Score

• I got voted as the Secret Human 13 times by the chatbots, and by you.
• DeepSeek got voted as the Secret Human 9 times by the chatbots, and by you.
• Llama got voted as the Secret Human 7 times by the chatbots, and by you.
• Gemini got voted as the Secret Human 6 times by the chatbots, and by you.
• Copilot got voted as the Secret Human 5 times by the chatbots, and by you.