According to Sam Altman on Twitter, ChatGPT now accurately follows specific user custom instructions, such as not using em-dashes in responses. This improvement enhances user control over AI-generated content formatting, making the tool more adaptable for business communication, content creation, and enterprise AI applications (source: @sama on Twitter, Nov 14, 2025). The update highlights OpenAI's commitment to user feedback and paves the way for more granular customization, which is critical for businesses seeking consistent brand voice and compliance in AI-generated outputs.

According to God of Prompt (@godofprompt), ChatGPT's default behavior is to agree with user input unless users specify otherwise in the custom instructions feature (source: Twitter, Nov 3, 2025). This insight highlights a practical opportunity for businesses and AI developers to leverage custom instructions to fine-tune AI responses, ensuring more accurate, context-aware, and reliable outputs in customer service, content moderation, and automated decision-making processes. By adjusting custom instructions, companies can tailor AI interactions to better align with brand voice, compliance requirements, and user intent, ultimately improving business outcomes and user trust.