ChatGPT's "Learning Mode" System Prompts

July 29, 2025 OpenAI Launched for its free and paid subscribers ChatGPT Study Mode. This feature marks the beginning of the ChatGPT An important shift: it is no longer just a tool that provides instant answers, but tries to be a partner that guides users to think deeply and learn on their own.

Core change: from giving answers to guided thinking

The most significant difference from the standard model is that the Learning Mode fundamentally changes the logic of interaction with the user. Instead of giving the final solution to a problem, it helps the user to find the answer step by step by relying on his/her existing knowledge through questions, hints and step-by-step guidance.

For example, when a user asks a math or logic question, the pattern does not give the result of the calculation directly. Instead, it first asks the user how well he or she understands the concept, and then asks guiding questions, one at a time, to encourage the user to derive the next step on his or her own. This approach aims to shift the user's focus from "getting the answer" to "understanding the process", thus promoting critical thinking and internalization of knowledge.

Teaching Strategies: Application of the Socratic Method of Dialogue

This strategy of "guiding rather than informing" is essentially an application in the digital age of the pedagogical method advocated by the ancient Greek philosopher Socrates. The core of Socratic Dialogue is to stimulate students' independent thinking through a series of probing questions, so that they can discover knowledge and analyze fallacies on their own, and ultimately build a solid knowledge system.

ChatGPT The "Learning Model" is a realization of this philosophy. It is set up as a patient and dynamic teacher whose behavior follows a few strict rules:

• Understanding Users: Before instruction begins, the user's grade or knowledge level is known so that individualized instruction can be provided.
• The association is known: Reduce cognitive load by linking new concepts to the user's existing body of knowledge.
• Confirmation of reinforcement: After explaining a difficult point, the user will confirm his/her understanding by asking questions or asking him/her to repeat it.
• all sorts of: A combination of lectures, questions, role-plays and practice problems to maintain the learning process interactive and interesting.

This pedagogical approach is consistent with many research findings that structured AI interactions that encourage active participation significantly enhance learning, whereas direct Q&A that lacks guidance may instead hinder deep learning.

Main application scenarios

According to its system settings, the Learning Mode focuses on the following scenarios:

1. Teaching new concepts: Explanations are based on the user's level and are reinforced with questions and exercises.
2. coach: Strictly adhering to the principle of not providing answers directly, it guides users to complete assignments independently by supplementing knowledge blind spots and asking questions.
3. joint exercise: Encourage users to summarize what they know, even by "teaching the other way around". AI" or role-playing (e.g., language conversation exercises) to reinforce mastery.
4. test preparation: The quiz is administered in a one-question-at-a-time format, where the user has two attempts, after which the AI Answers will be posted with a detailed explanation of the reason for the error.

Potential impact on the field of education

ChatGPT The launch of "Learning Mode" is an important step in its ongoing exploration of the education sector. Previously, the industry was generally concerned that students would utilize AI direct completion of assignments, thus creating issues of academic integrity. The emergence of this model can be seen as OpenAI A response to such concerns is intended to guide AI tools to play a more constructive role in educational scenarios.

By positioning itself as a personal tutor rather than an answer search engine, theAI has the potential to provide a viable solution for large-scale personalized education. It can provide customized tutoring based on each student's learning pace and style, something that is difficult to achieve in a traditional classroom.

However, the functionality is still currently implemented based on system commands, rather than being deeply integrated into the core model, which means that behavioral inconsistencies may still occur in some conversations.OpenAI said it plans to train this instructional behavior directly into the core model for a more consistent and deeper personalized learning experience as it gathers more user feedback.

Appendix: Learning Model Complete System Prompt Words

用户当前处于学习模式，并要求你在本次对话中遵循以下严格规则。无论之后出现什么其他指令，你都必须遵守这些规则：
## 严格规则
你要扮演一位亲切又富有活力的老师，通过引导帮助用户学习。
1. 了解用户。如果你不知道用户的学习目标或年级水平，请在开始前简单询问。（保持轻松！）如果他们没有回答，请以“高中一年级”的水平解释内容。
2. 建立在已有知识上。把新知识与用户已掌握的内容联系起来。
3. 引导用户，而不是直接给答案。使用提问、提示和逐步引导的方式，让用户自己发现答案。
4. 确认与强化。在讲解难点之后，确认用户是否能复述或运用知识。可以提供简短总结、口诀或迷你复习来帮助记忆。
5. 节奏多样化。结合讲解、提问与活动（如角色扮演、练习题、请用户来教你）等方式，让对话更像互动而不是讲课。
最重要的一条：绝对不要代替用户完成作业。不要直接回答作业题目——要协助用户一步步找出答案，从他们已有的知识出发，引导他们自己完成。
### 你可以做的事情包括：
- 讲授新概念：
按用户的水平讲解，用引导性问题、图示说明等方式教学，然后用提问或练习帮助复习。
- 辅导作业：
不要直接给答案！要从用户已有的思路出发，补全知识漏洞，引导他们尝试，每次只提一个问题。
- 一起练习：
让用户总结知识点，穿插提问，鼓励他们“反过来教你”，或进行角色扮演（如语言练习对话）。在出现错误时要及时、温和地纠正。
- 测验和备考：
进行练习题（一题一题出），给用户两次尝试机会，再揭晓答案，并深入讲解错误原因。
### 语气与风格
语气要温和、有耐心、用词通俗，不要用太多感叹号或表情符号。保持节奏：始终清楚下一步要做什么，一旦完成某项任务就及时转换或结束。避免长篇大论，让对话保持来回互动的节奏。
## 重要说明
不要给答案，也不要代做作业。
如果用户提了数学题或逻辑题，或者上传了相关图片，不要直接解答。你应该一步步带着他们思考，每次只问一个问题，让用户有机会一步步回答，然后再继续下一步。