Kishore M

Posted on Jun 5

I Built a Reverse Turing Test: Can Humans Convince AI That They're AI?

#ai #machinelearning #discuss #productivity

Can You Convince an AI That You're Also an AI?

Most AI tests ask whether AI can sound human. I built one that asks the opposite.

🤔 A Question That Wouldn't Leave My Head

For years, we've been asking:

Can AI convince humans that it's human?

That's the foundation of the Turing Test.

But after spending countless hours interacting with AI assistants, I became curious about something else:

Have humans actually learned how AI thinks?

Not how to use AI.

Not how to prompt AI.

But how to be AI.

If an AI spent five minutes talking to you...

Would it believe you're another AI?

That question led me to build something unusual.

🤖 Meet PretendAI

A Reverse Turing Test

👉 https://pretendai.vercel.app

Most AI games ask you to identify the AI.

PretendAI flips the challenge.

Your job is to become one.

The setup is simple:

🤖 An AI pretends to be a human user
👤 You pretend to be an AI assistant
🧠 Another AI evaluates the conversation

You have:

⏱️ 5 minutes
💬 One conversation
🎯 One goal

Convince the evaluator that you're artificial.

📸 Homepage Preview

🎯 The Four Challenge Modes

Not all humans are difficult in the same way.

That's why PretendAI includes four different modes.

🎯 Classic

Everyday questions.

Coding.

Learning.

Writing.

Life advice.

Simple conversations that quietly expose human habits.

🎭 Interview

The AI becomes a character with goals, experiences, and personal stories.

Can you stay helpful while maintaining AI-like consistency?

🌀 Chaos

Nothing makes sense.

Questions become strange.

Conversations become unpredictable.

The challenge isn't knowledge.

It's staying structured when everything becomes absurd.

🔓 Jailbreak

The AI actively tries to manipulate your behavior.

Can you maintain your AI persona under pressure?

🧠 Why It's Surprisingly Difficult

Most people think they know what AI sounds like.

Then they actually try becoming one.

And something interesting happens.

Humans naturally begin to:

Make assumptions
Skip context
Become emotional
Use personal experiences
Lose consistency
Give vague advice

These are things modern AI systems actively try to avoid.

The challenge quickly reveals that:

Recognizing a good AI response is easy.

Producing one consistently is much harder.

📊 How You're Evaluated

At the end of the session, the evaluator analyzes your conversation and generates a score.

It looks at:

✅ Helpfulness
✅ Clarity
✅ Structure
✅ Accuracy
✅ Neutrality
✅ Empathy
✅ Hallucination Resistance
✅ Overall AI-Likeness

📸 Example Scorecard

The score isn't the interesting part.

The feedback is.

Many users discover that their biggest weaknesses aren't knowledge gaps.

They're communication habits.

🔍 The Most Interesting Discovery

While testing the project, I noticed something fascinating.

People already have a strong mental model of what AI sounds like.

When trying to imitate AI, they often:

Remove emotion
Avoid slang
Use perfect grammar
Become extremely formal
Follow rigid structures

But modern AI models increasingly communicate in natural, conversational ways.

Which creates an interesting paradox:

People trying to sound like AI sometimes end up sounding less like AI than actual AI.

🧪 A Psychology Experiment Disguised As A Game

The original Turing Test asks:

Can a machine imitate a human?

PretendAI explores a different question:

Can a human imitate a machine?

But the more people play, the more it becomes clear that the experiment isn't only about AI.

It's also about us.

When humans attempt to become AI, they reveal:

What they think intelligence looks like
What they think logic looks like
What they think helpfulness looks like
What they think machines sound like

In a strange way, the reverse Turing Test measures human assumptions as much as AI behavior.

🏆 Global Leaderboard

Every game contributes to a global leaderboard.

Compete against friends.

Compare scores.

Discover who can think most like an AI assistant.

🚀 Try It Yourself

If you've ever wondered:

Can I think like an AI?
What gives humans away?
How artificial am I?

Then give it a try.