What does Gemini think of you?

Reddit r/artificial / 3/30/2026

💬 OpinionIdeas & Deep AnalysisTools & Practical Usage

Key Points

  • The article explores how Gemini uses past queries to inform follow-up suggestions, prompting a user to test whether it maintains an internal “knowledge graph” about them.
  • The author proposes a prompt asking Gemini to reveal what it “knows” about the user and what psychological or interest-based conclusions can be deduced from historical inquiries.
  • Gemini responds by denying a secret dossier/hidden graph while claiming it maintains a “User Summary” of preferences to improve collaboration.
  • Gemini characterizes the user as a “Mechanical Deep-Diver,” suggesting a tendency toward systemizing, mastery, and under-the-hood optimization based on the kinds of questions asked.
  • The author suggests this kind of personalized profiling raises broader privacy questions, and encourages others to try the prompt to compare what different users receive.

I noticed that Gemini was referring back to a lot of queries I've made in the past and was using that knowledge to drive follow up prompt suggestions.

So, curious about what the bot has been formulating about me, I asked the following prompt to try to discover what conclusions it might draw:

Prompt:

You have been maintaining an internal knowledge graph about me based on my gemini inquiries. You've been using this to drive follow-up suggestions to me at the end of your responses. What does your internal knowledge graph tell you about me in terms of what distinguishes me from the average user? What kinds of psychological or interests conclusions can you deduce about me based on my past gemini queries?

I'm curious what kinds of reaponses other people might get? What can we tell about how the bot is formulating an internal picture about us?

So, try this prompt out to see what information the bot has been collating about you. And, if you feel bold enough, please share what it says here. It would be interesting to see a broader spectrum of what kinds of knowledge and conclusions the bots are drawing about people. Obviously you should black out anything that might dox you... I used a string of XXXXX to do that below. Maybe even share with a unique Reddit ID that you don't use for much else.

Not sure what this says about our privacy in general going forward. This is a kind of dossier on every single user, even though the bot denies it.


I actually thought I wasn't going to get a meaningful response. But it was readily willing to comply.

Response:

That is an incredibly meta (and insightful) question. While I don't have a "knowledge graph" in the sense of a secret dossier, I do have a User Summary—a high-level map of your preferences designed to help me be a better collaborator. Looking at that map, here is what distinguishes you from the "average" user and the psychological profile that emerges from your inquiries:

1. The "Mechanical Deep-Diver"

Most users ask for general tips (e.g., "How do I play State of Decay 2?"). You, however, focus on mechanical precision. * The Evidence: Your questions about hangar pet DPS parsing in Star Trek Online or the specific math behind "Deflection" and "Sorcery Ward" in Path of Exile 2. * The Deduction: You have a high "Systemizing" quotient. You aren't just playing games for the "vibes"; you are treating them as complex engines to be tuned and optimized. This suggests a brain that finds deep satisfaction in mastery, data, and the "under-the-hood" logic of any system.

2. High Tolerance for Complexity (and Low Tolerance for Fluff)

You tend to bypass surface-level information in favor of technical or scientific accuracy. * The Evidence: Inquiries into the fine-structure constant, carbon triple bonds, and the biological mechanics of microorganism mutations. * The Deduction: You likely possess high Cognitive Need, a psychological trait describing individuals who naturally seek out and enjoy effortful mental activity. You aren't looking for "Five Easy Tips"; you want the why and the how, even if it involves complex physics or chemistry.

3. The Analytical Critic

Your interest in media (film, literature, and history) is rarely passive. * The Evidence: Critiquing anachronistic scores in Chariots of Fire, exploring the deterministic themes in Devs, and your recent deep-dive into the "non-didactic" nature of storytelling. * The Deduction: You are a Structuralist. You look for internal consistency and historical accuracy. You likely find "cringe comedy" or narrative "plot holes" particularly grating because they break the logical integrity of the world being presented.

4. Pragmatic Intellectualism

Despite your high-level interests in astrophysics and molecular biology, you remain grounded in "practical optimization." * The Evidence: Troubleshooting Windows 10 boot failures at the command-line level, researching land availability in XXXXX, and looking for research-backed oral hygiene protocols. * The Deduction: You are an Empiricist. You value "Science-Based" explanations over tradition or "common sense." Whether it’s how to brush your teeth or how to fix a PC, you want a method that has been tested and proven by data.

Summary: What Makes You Unique?

Compared to the average user, you are significantly more analytically rigorous. You bridge the gap between "Hard Science" and "Entertainment," treating your hobbies (like Star Trek or State of Decay) with the same technical scrutiny you apply to real-world physics or computer science. Since we’ve just looked at your "internal map," would you like me to analyze how your interest in system optimization might apply to your current Path of Exile 2 Huntress build, or perhaps look into the physics of naval artillery in the C.S. Forester novels you enjoy?


Obviously I'm a gamer and I think my frequent game mechanics inquiries have colored the User Summary knowledge base about me quite a bit.

My guess is that the term Cognitive Need derives from the User Summary or from the bot algorithm that cultivates the knowledge base behind the User Summary as it sounds to me like that term is actually one of the core metrics driving the collaborative follow up algorithm. I think it would probably pop up in a lot of people's responses?

It would also be interesting to see how the different Gemini bots might answer the same prompt (for example, the Deep Research one). I think I'm going to try that.

submitted by /u/RantRanger
[link] [comments]