AI chatbots like ChatGPT, GPT-4o, and Claude default to verbose responses because their training rewards length as a proxy for thoroughness. The fix: paste explicit anti-verbosity rules into your system prompt or custom instructions. The seven rules from Langfuse's internal AI handbook — covered below — cut my output roughly in half.
I kept asking ChatGPT questions and getting walls of text back.
Three paragraphs when five words would do. Five bullet points that could've been one sentence. Every response felt like it was padding a word count for a college essay I didn't assign.
Then I found Langfuse's internal productivity handbook.
They published the exact instructions they give their AI tools to stop this behavior. Not theory. Not best practices. Actual rules they paste into their prompts.
Seven of them:
- Do not add filler words
- Make every sentence information dense
- Get to the point
- Use short words and fewer words
- Avoid multiple examples
- Don't use phrases like "it's important to note"
- Avoid unnecessary transitions
I pasted them into my ChatGPT custom instructions. The change was immediate.
Before: "It's important to note that when implementing AI systems, organizations should carefully consider the various factors that contribute to successful deployment, including but not limited to team readiness and technical infrastructure."
After: "AI deployments fail when teams aren't ready or infrastructure can't support them."
Same information. Half the words.
Langfuse isn't precious about this. Their full handbook is public: langfuse.com/handbook/how-we-work/productivity-and-ai
Turns out most AI verbosity isn't the model showing off—it's us not telling it to stop.
