Anyone Have a Working System Prompt to Stop GPT-5 from Using Dashes and Saying “Nice” All the Time?

Effective Strategies for Improving GPT-5 Response Formatting and Tone: Addressing Common Challenges

As AI language models like GPT-5 continue to evolve and become integral tools for communication and content creation, users often encounter recurring issues with the quality and tone of generated responses. Two common challenges include the inconsistent use of punctuation—specifically dashes—and an overly enthusiastic, sometimes out-of-place, conversational tone such as frequent expressions of approval like “Nice!” or “That’s great!”

Understanding and managing these behaviors is crucial for professionals seeking to integrate AI seamlessly into their workflows, whether for drafting formal documents, technical content, or professional correspondence. Below, we explore strategies to optimize AI responses by adjusting prompt engineering techniques—especially system prompts—to achieve more appropriate formatting and tone.

1. Addressing Unnecessary Use of Dashes

GPT-5 may employ dashes in responses for various reasons, often defaulting to a more informal or conversational style. To mitigate this:

• Explicitly specify in your prompt that you prefer responses without dash-separated thoughts or sentences.
• Use directives such as, “Please avoid using hyphens or dashes in your responses unless absolutely necessary,” to guide the model’s formatting choices.

• Incorporate instructions about preferred punctuation and structure to set clear expectations upfront.

• Moderating Overly Cheerful or Redundant Expressions

The habitual use of praise or cheerfulness like “Nice!” at the beginning of replies can undermine professionalism, especially in serious contexts. To curb this:

• Clarify in your prompt that the tone should be neutral, formal, or aligned with the desired professionalism level.
• For example, include statements such as, “Please respond in a professional tone without unnecessary affirmations or expressions of excitement.”

• Consider instructing the AI to adopt a specific tone variation (e.g., “Maintain a neutral, objective tone throughout your responses”).

• Crafting Effective System Prompts

System prompts serve as foundational instructions that significantly influence the behavior of AI models. To design a prompt that discourages unwanted behaviors:

• Combine detailed tone and formatting instructions into a comprehensive system prompt. For instance:

“You are an AI assistant providing clear, concise, and professional responses. Avoid using dashes unless necessary, and do not include casual or enthusiastic expressions such as ‘Nice!’, ‘Great!’, or similar phrases at the beginning of your replies. Maintain a neutral tone suitable for professional or serious contexts.”

• Test and refine your prompt based on the AI’s output, adjusting instructions for clarity