You’ve probably seen the meme:
"Ask ChatGPT to pick a number between 1 and 50 — it always says 27."

I wanted to find out if that was really true, even when done at scale.

So I asked the same question over 800 times across ChatGPT, Perplexity, Gemini, and Copilot using a tool I am building called Radix AI.
I changed phrasing, location, and tone to simulate real variation.

You can view the data report here on this looker studio.

Results:

  • 27 was the most common answer (~60% of the time)
  • But 37, 42, and even Python code appeared regularly
  • ChatGPT gave me 16+ different responses based on how I phrased the question
  • Some models used web sources (Reddit, blogs); others didn’t

Why these results:

  • 27 & 37 are statistically common “random” picks in human behavior (LLMs reflect that)
  • 42 comes from pop culture (Hitchhiker’s Guide to the Galaxy)
  • Python code showed up when the prompt included words like “generate”. Thanks to 11th grade CS assignments across the world.

I used Radix AI to collect data, google sheets to clean and looker studio to visualise.

Posted by Agitated-Arm-3181

9 comments

  3. Didn’t know 27 was a common pick for people, it’s a cube so it doesn’t strike me as “random”

    I know 7 is the most common answer for single-digit numbers, because it’s the biggest prime there, it’s awkward to handle in everyday maths

  5. Gemini gave me 47, and this is its thinking part:

    **Determining Randomness Generation**

    I’ve determined that the best approach is to leverage a programming language, specifically Python, to handle the user’s request for a random number. The provided parameters seem sufficient, so now I’m focused on the implementation details.

    **Developing Python Implementation**

    I’ve crafted the Python script, incorporating the `random` module and the `randint()` function to produce the random integer. The parameters, namely the range (1 to 50), are now integrated within the code. Currently, I’m verifying the script’s functionality to ensure it accurately generates and returns a random natural number. Next, I plan to finalize the user-facing output presentation.

    **Determining a Random Number**

    I’ve generated a random integer between 1 and 50 using Python’s `random.randint()` function, as requested. The outcome was 47, which confirms the successful execution. Next, I’ll consider how this result might be integrated into a larger process.

    **Confirming Number Generation**

    I’ve successfully obtained the random number 47 within the specified range, which the function did correctly. This has now been stored and is ready for use in the next stage

  6. I asked Mistral and got 23, then asked the same question again, but with “thinking” on but still got 23 even though it made a python script, so I challenged it and the self-prompt was kind of interesting. It challenged why it happened and concluded it was just random chance. It redid the algorithm and gave 16.

  8. > 27 & 37 are statistically common “random” picks in human behavior

    Got a real citation for that?

Comments are closed.