This projects consists of 2 parts: the game itself and a companion database app for AI-generating new articles.

Can you spot which text was written by a large language model? In this game, you're shown two articles: one generated by a variety of different AI models and one from Wikipedia!

In today's world where you're constantly surrounded by AI slop, it's important to be able to distinguish it from real human-written text! Train your ability to spot AI with this game!

I recommend this amazing and exhaustive Wikipedia article on the signs of AI writing, which actually inspired me to make this game.

The article titles that are currently in the game are manually picked to be generally recognizable and well-documented. The number of articles is constantly expanding as I add more. If you want to suggest an article title yourself, you can do that here!

A companion app (located in the /database folder) helps to create new entries and edit the database. It's responsible for saving the Wikipedia article's "summary" paragraph (first paragraph) and interacting with an API to make a AI-generated counterpart..

The currently used prompt for the LLM is as follows:

"Generate a summary of the object provided in the input, basically like the summary section of a Wikipedia page. Your repsponse should be concise. Do not include links, bulletpoints, or markdown formatting. Do not encase words in asterisks, this does not make the text bold. Respond in about 70-120 words." (note: the prompted preffered word count may change to better match the Wikipedia summary length)

This prompt is usually able to generate a similar summary to the one given on Wikipedia quite well.

• Progress bar (leveling system)
• Log the time it takes to respond
• Add a "loading" disclaimer
• Make landing page
• Ability to turn off typing animation
• Display a "consensus" percentage (e.g. 70% picked the same answer as you)
• Collect, aggregate and display statistics

AI (specifically GitHub Copilot) was also used in the creation of this project! However, I find most of the code written by Copilot a bit clunky and unnecessarily complicated, so most of the time I rewrite it myself and use Copilot just to give me an idea.

A part of the database companion app (listing and deleting items from the DB) is written fully by Copilot. This was done to focus more on the main part of this project -- the game itself.

This project was created as part of Hack Club's Summer of Making 2025. Many thanks to Hack Club for such an amazing event 😊!