ChatGPT has a major update and officially launched the Study Mode: it no longer just gives you answers, but guides you to think about the problem step by step like a teacher. This feature is already available to Free, Plus, Pro and Team users, and ChatGPT Edu users will also be able to use it in the coming weeks.

In addition to simply tutoring a question, it can also be used to prepare for an exam.


Leah Belsky, vice president of education at OpenAI, pointed out: When ChatGPT is used for teaching or tutoring, it can significantly improve student learning. But if you treat it just as an "answer machine", it will hinder learning.

This is true. Data shows that about one-third of college students are using ChatGPT to assist with their studies. However, schools and parents have always been worried that students will use it to copy answers directly, damaging their critical thinking skills.

For example, a previous MIT study found that people who used ChatGPT to write essays had lower brain activity during the writing process than people who used Google search or did not use ChatGPT at all.


Some netizens suggested that students can switch to normal mode at any time and continue copying answers.


However, OpenAI revealed that it will not currently launch a tool that allows parents or teachers to lock students in student mode, but this feature may be explored in the future.

Some people lamented that another 1,000 GPT shell projects were finished within one day.


How the ChatGPT learning model is built

The biggest feature of the learning mode is that it will not tell you the answer directly.

When you ask a complex homework question or want to understand a new concept, ChatGPT will first ask guiding questions to let you think about the steps to solve the problem, or summarize the concept in your own words.


The design of the learning model does not come out of thin air, but is the result of in-depth cooperation between OpenAI and teachers, scientists and pedagogy experts.

Its underlying technology is driven by a customized set of system instructions that incorporate long-term research in learning science and are designed to promote deep learning in users, including encouraging active participation, managing cognitive load, developing metacognition and self-reflection, cultivating curiosity, and providing actionable and supportive feedback.

Key features include:

Interactive Questioning: A combination of Socratic questioning, prompts, and self-reflection prompts to guide understanding and promote active learning rather than providing direct answers.

Scaffolded responses: Information is organized into easy-to-understand sections, highlighting key connections between topics, keeping information engaging with appropriate context, and making complex topics less confusing.

Knowledge Checks: Quizzes and open-ended questions, along with personalized feedback to track progress, support knowledge retention and the ability to apply knowledge in new contexts.

Even more considerately, the learning mode will provide customized teaching content based on your skill level and memory of past conversations. Complex information will be broken down into easy-to-understand modules, highlighting the connections between knowledge points, which not only ensures sufficient contextual information but also avoids information overload.

Users can also turn on or off the learning mode at any time during the conversation, completely adjusting it according to their own learning goals.

Let’s look at a practical case:

In the demonstration, when students want to learn "Game Theory", the learning model first develops a multi-stage learning roadmap from core basics to practical applications.

The teaching process will be interspersed with guiding examples, thought experiments, and continuous questions to test understanding.


It will be trained into the model in the future

OpenAI said that the current learning model is only "the first step in a long journey."

This function is now implemented through custom system prompt words. OpenAI chose this method to quickly learn from real student feedback and improve the experience.

But this can lead to some inconsistent behavior and errors in conversations, and they plan to train this behavior directly into the main model in the future after learning what works best through iteration and student feedback.

Future improvements include clearer visual representations of complex or text-heavy concepts, goal setting and progress tracking across conversations, and deeper personalization for each student's skill level and goals.

OpenAI is also conducting further research with multiple partners through the NextGenAI program and working with the Stanford Learning Accelerator’s SCALE program to study how AI tools can impact learning outcomes in areas such as K-12 education.

OpenAI promises to publish in-depth analysis of the connection between model design and cognition, and to work with the broader education ecosystem to ensure that AI can benefit learners around the world.

It is worth noting that OpenAI CEO Altman himself is skeptical of traditional education. He once said that his children "probably will not" go to college. After all, he himself is halfway through.

Altman also predicted that education will look completely different in 18 years.


This view adds a deeper strategic intention to OpenAI’s layout in the education field. They may want to fundamentally reshape the future education model.

Reference links:

[1]https://openai.com/index/chatgpt-study-mode/

[2]https://techcrunch.com/2025/07/29/openai-launches-study-mode-in-chatgpt/