Serious Issues with Accuracy – Repeated Character Count Failures

ChatGPT
1

Hi OpenAI,

I’ve been using ChatGPT for important work and have recently hit a serious issue that’s costing me hours of time and undermining my trust in the product.

Over the past few days, I’ve asked the model—multiple times—to write a LinkedIn message under 750 characters and tell me the exact character count. Not once did it give me a correct count. Every single time, it told me the message was 749 or something similar—but when I checked myself, the counts were far lower (e.g. 571, 484, 516).

Even after pointing this out repeatedly, the model kept returning incorrect numbers, claiming it had “verified” them. This wasn’t an occasional slip—it was every single attempt. I lost hours trying to fix and re-check what should have been a simple task.

This might seem small, but it highlights a much bigger issue: if I can’t trust the model to count characters accurately, how can I trust it on anything more complex? I’ve paid for this service and expected basic reliability. Right now, it feels broken.

Please let me know:

  • Is this being tracked as a known issue?
  • Has anything changed in how the model handles specific requests like this?
  • What can be done to prevent this in the future?

Stephen Johnston

2

Hi welcome back!

ChatGPT cannot calculate characters in chat. Yo can use python to control it.

Try this prompt:

You are a professional assistant helping me write optimized LinkedIn messages. Please do the following:

1. Write a professional LinkedIn message that:
   - References a recent post about AI or technology
   - Briefly states that I'm focused on ethical AI implementation in enterprise environments
   - Invites the recipient to connect, exchange ideas, or explore collaboration
   - Uses natural, persuasive tone suitable for networking

2. The message must be under 750 characters, and ideally close to 750 without going over. Do not include placeholders like "[Name]".

3. After generating the message, use Python (Advanced Data Analysis tool) to:
   - Print the exact character count using the len() function
   - Show only the number, not an estimated value

Important:
- You must use the Python tool to calculate the character count, not simulate it.
- Do not include quotation marks or newlines in the count.

Return both the message and the verified character count.