If you’ve ever struggled to tell if the customer service rep you’re chatting with is an AI or real human, you’re not alone. Many people have willingly engaged themselves in the task of making this difference in a game, Human or Not.
Human or Not is not like the typical game. It is a budding pastime that tests and reveals the player’s abilities to differentiate a chatbot from a human. However, Human or Not stretches beyond this simple summary.
In this article, you will learn about the details of Human or Not, how it works and the ultimate purposes of the game.
Also read: Top 10 AI Games of 2025
The Turing Test
Human or Not has its roots in the Turing test. This test is a concept introduced by a British Mathematician and Computer Scientist. Alan Turing. Turing was riddled with the question: Can machines think?
His search for an answer to this took an, at that time, unconventional route. Instead of trying to define think, he proposed a practical experiment dubbed the ‘Imitation Game.’ In the setup, a human is tasked with communicating with two anonymous participants, one human and one machine, and telling which is which.
If the human cannot reliably label the participants, the machine is said to have passed the test. Today, this social experiment still occurs, but on a much larger scale through “Human or Not.”
What is Human or Not?
Human or Not is an online social experiment created by AI21 Labs. Its purpose is to ascertain how well people can distinguish between human and conversational AI, no matter the sophistication.
The groundwork, based on the Turing test, has been tweaked and reworked to a more playful, fast-paced format. And anyone with an internet connection can join.
How It Works
A game session lasts 2 minutes, after which the player makes a submission. At the beginning of a session, the system randomly pairs the player off with an AI chatbot or a human. The chatbot runs on an advanced AI model (AI21’s Jurassic 2), which is capable of generating natural, contextually appropriate responses.
The player can decide to talk about anything ranging from casual banter to deeper topics that reveal a thought process or a lack thereof. In this exchange, there is no voice, video, or profile information sharing. This cuts out confounding factors that can influence the players’ judgment.
At the end of this two-minute interaction, the player picks between the two options. The guess is recorded alongside its accuracy.
A fascinating aspect of Human or Not is the various strategies players employ to make guesses. Some ask personal questions or time-specific events in the hopes that an AI might struggle to show glimpses of a real personality.
Others have resorted to linguistic tasks. They could ask their chat partner to spell something backwards, change the language, or intentionally make errors. However, these strategies may not always be reliable as advanced AI has been trained to mimic human errors, slang, and occasional mistakes.
The Results
The data has been telling. It has revealed the strengths and weaknesses of human intuition. Overall, players’ guesses were right 68% of the time. People made more accurate guesses (73%) when speaking to a human than when speaking to an AI (60%).
The Goal
The primary goal of Human or Not is to determine the ability of people to distinguish between humans and AI. It teaches players to think critically about communication and spot subtle patterns.
From the creators’ standpoint, Human or Not is categorically one of the largest public Turing-style experiments ever conducted. It has helped to gather massive data on human-AI interaction, which serves secondary purposes:
1. Evaluate AI’s Language
Human or Not occurs in unhinged and uncontrolled settings filled with humour, slang, and taboo language. This contrasts with the strict, lab-like environments where the prompts and answers are structured and held to a set standard. This allows researchers to clarify AI’s performance, abilities, and limitations. They then use the limitations to improve future models.
2. Observe Human Detection
The game provides significant insight into the methods people deploy to “catch AI”. It uncovers the clues that make AI appear robotic and the ones that pass it off as human. This knowledge can be applied towards designing transparency or safety features that let users know when they’re talking to a machine.
3. Statistical Data
AI’s full impact on and interactions with humans as a race are yet to be fully understood. From a demographic standpoint, it can help to identify certain groups that are better at spotting AI. It can also influence ethics and regulations surrounding AI.
FAQs
1. How Does the Human or Not Game Work?
In Human or Not, players are tasked with finding out whether their chat companion is an AI chatbot or a real human. Over 2 minutes, there is a text exchange between the player and an anonymous character.
2. What’s the Point of Human or Not?
The point of Human or Not is to see if you can tell the difference between talking to a human or an AI in a short, text-based chat. It’s a game for players, but for the creators, it’s also a large-scale experiment to study how realistic AI conversations can be and how people try to spot bots.
3. Is Human or Not a Safe Site?
Yes. Human or Not is safe to use. The game runs in a browser, doesn’t require personal details, and only shares the messages between you and your chat partner during the round. However, you should still avoid sharing sensitive or private information in any online chat.
4. Who Will Win, AI or Human?
It depends on the conversation. In past results, humans guessed correctly about 68% of the time. They were better at spotting humans than bots, but AI still fooled people often enough to make the game challenging. Sometimes the human wins, sometimes the AI does.
