When students submit questions to chatbots, the most straightforward outcome is to receive an answer; however, what teachers truly want to know is where exactly the students get stuck. On October 15th, Teachers' Day, OpenAI announced an educational update, introducing a Brazilian Portuguese version of the OpenAI Academy course series ("AI for Educators"), and once again demonstrated how ChatGPT enables learners to reason step by step through questioning, hints, and checks. It's important to distinguish that what is newly launched this time is merely a language version of the course, not the Study Mode which was first released on that day, nor is it a clinical approach that has been proven to improve all students' grades.
According to the official introduction, this free course is aimed at primary school, secondary school, and higher education teachers. It covers various aspects such as lesson preparation, designing activities and assessments, and writing communication materials. At the same time, it emphasizes that teachers should retain their own teaching judgment. Launching the course in a different language may seem like a simple translation on the surface, but in reality, it can affect whether teachers are able to integrate these tools into their local classrooms. Teaching methods, communication with parents, and homework evaluation vary from region to region. If the English examples are directly copied, teachers may learn how to use the tools, but they may not understand when it is not appropriate to do so. Whether the course can truly be effective still depends on the localization of the content, the time spent on teacher training, and feedback from the classrooms, rather than just the announcement of its launch.
AI can break down problems, but it cannot think on behalf of students.
The basic design of Study Mode is a conversational approach: it first asks the user what they know and what they want to learn, then progresses through prompts, explanations, and checks to ensure understanding. OpenAI provides an example from a biology class reviewing enzymes: students can let the system ask one question at a time, and after answering it themselves, they can then indicate the areas that need review. This kind of interaction is different from simply “giving me the correct answer right away”; it allows students’ thought processes to be revealed. However, whether the questions posed by the system are appropriate and whether it identifies conceptual misunderstandings still requires judgment from the teacher; an interface that continuously asks questions does not automatically equate to good teaching.
One of the most challenging aspects in the classroom is distinguishing between students who are temporarily unable to understand, those with weak memory, and those who seem to know how to do it but actually don't comprehend it. For example, when learning about evaporation, students may be able to recite that "liquids turn into gases," but they may not be able to explain why the clothes on a clothesline dry. OpenAI suggests that teachers should first have students use Study Mode to explore the concept on their own, and then have them write down their own explanations and raise questions for class discussion without relying on any aids. This step of "leaving behind aids" is crucial: if the final expressions are still written by AI, what teachers see is merely the system's ability to process language, not the students' actual understanding. Only by treating the generated explanations as practice material, rather than directly as learning outcomes, can the true meaning of teaching assessment be preserved.
For teachers, AI can help draft activities and quizzes, but it can also easily result in a bunch of materials that seem complete yet are actually not suitable for the class. The difficulty of the questions, local curriculum objectives, students' reading levels, accessibility requirements, and grading standards cannot all be addressed with a set of generic prompts. A more prudent approach is to have the teacher determine the learning objectives first, and then use the AI draft as material to be edited: remove claims without evidence, check if the answers are unique, try out the questions, and estimate the actual time required in class. If a course can teach these review processes, it is more valuable than just presenting dozens of “efficient prompts.”
When evaluating such tools, don't just count the number of logins and questions asked.
When technology companies release educational products, what they often focus on first is usage: how many students activate the features and how many questions they ask. However, schools need to ask a different set of questions: whether students are willing to try it out on their own, whether they can understand the material independently without the aid of AI, whether teachers can more quickly identify misunderstandings, and whether students with different levels of knowledge will benefit from it. To assess learning outcomes, it is necessary to compare before and after use, conduct long-term tracking, and have appropriate control groups. This announcement OpenAI does not provide data on academic performance improvements resulting from the Brazilian Portuguese course, nor does it draw the conclusion that Study Mode is effective in all subjects. Reports should not make up for such missing information.
There are also constraints regarding privacy and fairness. Classroom conversations may contain information about students' homework, age, signs of learning difficulties, or family circumstances; when teachers choose to use any external services, they must consider school policies, parental consent, and data management practices. Even if the system is not used for formal exams, differences in internet access, devices, and teacher training resources can lead to varying outcomes when using the same tool in different schools. The real challenge in educational technology is not to make responses flow smoothly, but rather to enable teachers to clearly see what students have done and what they have not done, and to have the ability to switch back to face-to-face teaching at any time.
OpenAI This time, they also released a short video set in a university classroom, depicting the changing relationship between teachers and students and AI. However, an advertisement is a form of promotional material, not a randomized controlled experiment; one cannot infer the actual quality of learning from the classroom atmosphere depicted in the video. What is truly worth tracking are the specific teaching modules of the Portuguese courses, teachers' feedback, the usability of Study Mode in local classrooms, and how schools handle the boundaries for students to complete assignments independently.
If a school wishes to conduct a small-scale trial, they can start by selecting a unit that a teacher can personally review. Have students complete explanation, migration, and error-correction tasks both with and without the use of Study Mode. Then, observe whether they can articulate their thoughts independently. Such a comparison does not automatically constitute a rigorous experiment, but it is closer to the learning objectives than simply counting "how many questions were asked." Teachers can also use this to identify in which aspects the tool provides assistance and in which aspects it instead makes students dependent on hints.
What can be confirmed today is that a course for educators has added a Brazilian Portuguese version, and Study Mode continues to be promoted as a tool for step-by-step questioning. It provides teachers with a new entry point, but it does not replace their ability to ask questions, nor does it bypass the verification of learning outcomes. With answers coming more quickly, perhaps what is most worth protecting in the classroom is still that “why” that students need to articulate on their own.
