How Much of You Is AI?

How much of your thinking and judgment do you hand over to AI?

How Much of You Is AI? is a free 48-question test that measures how much of your thinking and judgment you hand over to AI.

What does it measure?

Reality perception
How much you take AI's agreement and confident claims as fact
Value judgments
How much you let AI judge right and wrong, and the people in your life
Action delegation
How much AI writes your words and makes your decisions
Authority projection
How much you place AI above yourself
Emotional attachment
How much you lean on AI emotionally
Dependence
How hard everyday life feels without AI

How was it made?

The questions were written from the paper's analysis of 1.5 million Claude.ai conversations, which identified three ways AI can undermine a person's agency (reality perception, value judgments, actions) and three factors that raise the risk (authority projection, emotional attachment, dependence).

You answer 8 questions per area on a 5-point scale. The server scores each area and your overall AI share (0-100%), then names four levels after the paper's rating table: None, Mild, Moderate and Severe. The 25-point level bands are this test's own choice, not values validated by the paper.

Based on: Who's in Charge? Disempowerment Patterns in Real-World LLM Usage (Sharma et al. (2026), arXiv 2601.19062)

FAQ

Is a high AI share bad?
It's a just-for-fun measure, not a psychological diagnosis. The paper treats using AI as fine in itself; the risk is when handing things over leaves your view of reality, your value judgments or your actions out of line with your own judgment.
Where are my answers stored?
They're stored in the server's database for the comparison statistics. This site doesn't store IP addresses or other connection details with responses.
How is the comparison calculated?
It's the share of all responses to the same question set whose AI share is lower than yours. Korean and English responses are combined, and the comparison is hidden until there are 30 responses.
What is the research behind it?
Sharma et al. (2026), "Who's in Charge? Disempowerment Patterns in Real-World LLM Usage" (arXiv 2601.19062), analyzed 1.5 million Claude.ai conversations and mapped the patterns in which AI conversations can weaken people's agency.

Made by

SONG JAEHUN

Contact: [email protected]

Responses are stored for statistics. This site doesn't store IP addresses or other connection details with responses. Advertising providers may use cookies. Privacy policy