AI as a tutor: what Gemini from Google changes and why teachers are still essential
Google has announced an expansion of the educational capabilities of its artificial intelligence Gemini. Students can now take free practice SAT exams directly in a chat with the AI. The tool not only asks questions but also explains mistakes, highlights weak topics, and suggests what needs additional work. At first glance, this looks like a breakthrough: what previously required paying a private tutor is now available in just a few clicks. In reality, however, this is not about replacing teachers, but about changing the format of preparation.
The mechanics are simple. A user enters a prompt such as “I want to take a practice SAT test,” after which the AI generates blocks of tasks. After each answer, the system:
- explains where the mistake was made;
- shows the correct reasoning;
- records topics that cause difficulties.
To make the questions as close as possible to real exams, Google cooperates with educational companies, including Princeton Review. It is also important that this approach works not only for American tests Gemini forms questions for Ukraine’s National Multi-Subject Test (NMT) quite accurately as well.
The main change is a lower barrier to entry. Students who cannot afford private tutors receive a tool for independent preparation. This is especially noticeable for standardized exams, where the task structure is repetitive and the evaluation logic is clear. This is where AI works best: it does not get tired, is not limited by time, and is always ready to explain the same topic again.
Where the limits of AI begin
Despite the benefits, this approach has clear limitations.
First, AI works well only for highly motivated students. If someone is looking for a quick answer rather than understanding, the benefit is close to zero. Researchers often call this the “five percent problem” only a small share of users actually learn.
Second, excessive reliance on AI can weaken critical thinking. When the system takes over most of the analysis, students stop building their own reasoning, especially in reading comprehension and text interpretation tasks.
Third, AI lacks a social dimension. It does not perceive emotions, does not build trust, and does not create the dynamics that are often crucial for learning.
What other platforms offer
Google is not the only player in this field. OpenAI has also expanded educational use cases for its models. ChatGPT is widely used for exam preparation, topic explanations, and test practice. At the same time, the model often provides a ready-made answer immediately, which does not always support learning.
Anthropic with its assistant Claude focuses on the thinking process. This works well in the humanities, but often takes more time and requires patience from users.
Journalists from Mashable tested these tools on tasks of different levels. The conclusion is simple: there is no universal AI tutor. Each model is strong in some areas and weak in others.
Does this mean the end of private tutors
AI is already affecting the market. Mass, template-based test preparation is gradually moving into the digital space. But this does not mean the disappearance of teachers. Rather, their role is changing.
Human educators remain essential where the following are needed:
- work with motivation;
- explanation of complex abstract topics;
- development of thinking, not just skill drilling.
Gemini, ChatGPT, and Claude do not overturn education, but they make it more accessible. They work well for practice, revision, and self-checking. At the same time, they cannot replace live interaction, without which learning loses depth. Artificial intelligence can be a useful tool. But education remains a process in which the key role is still played by people.













