73.7% of Japanese High School Students Use AI as Surveys Reveal Gaps in Learning and Guidance

AI Becomes an Everyday Study Companion

73.7 percent of Japanese high school students use conversational artificial intelligence tools such as ChatGPT, according to a survey released by the Gakken Research Institute for Learning and Education. The findings show how deeply AI has entered teenage life, helping students with homework, information searches and creative work while prompting concern about how much thinking they leave to the technology.

Gakken conducted the online survey in November 2025 among 2,400 students across Japan. It collected responses from 100 boys and 100 girls in each grade, from the first year of elementary school through the third year of high school. That design gave each grade and both genders equal representation within the survey.

High school students had the highest rate of use. Among junior high students, 43.2 percent used conversational AI, compared with 36.6 percent of elementary pupils. The gap between high school and elementary school was 37.1 percentage points, illustrating how much more common the technology becomes among older students.

ChatGPT and similar services respond to questions and instructions in conversational language. Students can use them to seek explanations, work through assignments, research a subject or develop ideas. Some AI tools also generate images, giving children a way to experiment with creative projects.

The adoption described in the surveys occurred less than three years after consumer AI tools became widely available. For many teenagers, AI is becoming an ordinary study resource alongside familiar tools such as calculators and textbooks. Its appeal suggests that students find useful support in areas where tutoring, access to information and opportunities for creative expression may be limited.

Parents and educators now face an immediate practical task: helping children use that support while continuing to think for themselves. Students across the age groups already have habits and preferences, so school policies must respond to existing behaviour.

Why the Surveys Give Different Numbers

Kankō Gakuseifuku, a school uniform manufacturer, surveyed 1,200 junior high and high school students in January 2026. Nearly 80 percent said they used AI tools frequently or occasionally. The result points to widespread use among teenagers and suggests that access and acceptance may be expanding.

The Children and Families Agency recorded a lower figure: 46.2 percent of high school students used generative AI online. Its survey was the first to track AI use since the agency began conducting internet studies in 2009. Gakken's corresponding figure for conversational AI was 73.7 percent.

Survey timing, the selection of respondents and the wording of questions about platforms and frequency may help explain the different results. Conversational AI describes tools that exchange messages with a user; generative AI is a broader term for technology that produces material such as text or images. What a questionnaire asks about can affect who answers yes.

The surveys nevertheless describe the same broad development: AI has moved from a relatively unfamiliar technology to a mainstream source of educational assistance within approximately two years of public availability. The findings span several dates, alongside research on universities:

• 2009: The Children and Families Agency began its series of internet studies.
• 2023: Researchers examined about 1.35 million syllabi from 707 universities and found 528 mentions of generative AI.
• 2024: A study of 1.24 million syllabi from 679 universities found 7,442 references.
• April 2025: A survey of first year medical students at Chiba University examined AI use and education.
• November 2025: Gakken surveyed 2,400 elementary, junior high and high school students online.
• January 2026: Kankō Gakuseifuku surveyed 1,200 junior high and high school students.

The sequence places students' growing use alongside changes in university teaching. By April 16, 2026, the central issue was already how to manage a technology widely used outside, and sometimes within, formal lessons.

Homework, Research and Creative Work

High school students who used conversational AI most often turned to it for academic help. In Gakken's findings, 42.3 percent cited studying and homework as their primary purpose, while 26.0 percent primarily used it to find information. Older students appear to regard AI chiefly as a practical resource for schoolwork.

That assistance can take several forms: clarifying a difficult concept, checking work or suggesting ideas for an assignment. Each involves a different relationship between the student and the tool. Asking for an explanation leaves room for the learner to work through the subject, while requesting a completed answer gives the technology more of the task.

Junior high students showed a closer balance between research and study. Information searches accounted for 17.8 percent, while studying and homework accounted for 17.7 percent. Their lower figures accompanied lower adoption in this age group, and the near equality suggests that their use spans both finding information and seeking academic assistance.

Elementary pupils showed the widest range of activities. Among their reported uses, 44.0 percent involved finding information and 32.6 percent involved studying. Another 23.7 percent used AI to create illustrations and images.

Image generation gives younger children a different route into the technology. A written instruction can become a picture, making AI a medium for artistic expression as well as a place to ask questions. Older students, facing more demanding coursework, appear to put greater weight on assistance with text and academic tasks.

The age differences matter for teaching. A child generating an illustration and a teenager seeking help with a calculation are using related technology for different purposes. Guidance needs to address what students actually do, including how they judge an answer, develop their own ideas and decide which parts of a task to complete independently.

Girls Report More Frequent Use and More Conversation

Girls in the Kankō Gakuseifuku survey reported more frequent AI use than boys. Among female students, 46.8 percent said they used the tools frequently, compared with 36.0 percent of male students. That is a gap of 10.8 percentage points.

The difference suggests that girls may have adopted the technology more readily or found more immediate value in its applications. The survey also showed a much larger gap in using AI for consultation or conversation: 49.9 percent of girls reported that activity, compared with 23.0 percent of boys.

Nearly half of the girls surveyed therefore used AI for a purpose extending beyond straightforward academic assistance. Their rate was more than twice the figure for boys, with a difference of 26.9 percentage points. For some students, a chatbot can serve as a place to seek advice, organise thoughts or discuss worries without feeling judged.

AI's role as a conversational companion brings social and emotional concerns into the education discussion. Parents and schools need to consider social isolation, adolescent mental health and whether conversations with a chatbot sit alongside human relationships or take their place. The survey figures describe the activity; the reasons individual students choose it remain unexplained.

What Students Say About Their Thinking

Gakken also asked students whether their thinking ability had become stronger or weaker after using generative AI. The most common answer in every age group was that they felt no particular change: 42.1 percent of elementary students, 49.8 percent of junior high students and 42.1 percent of high school students chose that response.

Elementary and junior high students more often reported stronger thinking than weaker thinking. Younger children may feel that AI extends what they can do or helps them understand material more quickly. Their responses suggest a sense of support or increased capability.

High school students showed the reverse pattern. More said their thinking had become weaker than said it had become stronger. As academic demands rise and AI becomes a more intensive part of study, older students may become more aware of dependence or feel that taking shortcuts reduces their own practice in solving problems.

The findings concern students' perceptions of their abilities. They capture how learners experience the technology, including the possibility that younger pupils feel helped while older students recognise that they are handing more work to AI.

Hiroyuki Masukawa, a professor of cognitive science at Aoyama Gakuin University, expressed concern about the findings. He called for attention to whether students' use supports educational development or mainly creates a feeling of greater efficiency. Describing the need to teach children how to use AI while developing their own thinking, he said:

It is essential to foster AI literacy and make sure that children can use the technology as an assistant to help improve their thinking ability.

AI literacy means understanding how to use the tools, question their answers and recognise their limitations. Masukawa's point centres on the learner's role: students need to remain involved in reasoning, even when a system can produce a quick response. The contrast between younger and older pupils makes that concern especially relevant as children move into more demanding schoolwork.

When Assistance Becomes Dependence

Approximately 20 percent of respondents in the Kankō Gakuseifuku survey relied entirely on AI to provide answers or perform calculations without using their own reasoning. For that group, the technology had moved beyond supporting a task to doing its central intellectual work.

Leaving answers and calculations to AI changes the student's opportunity to practise. A completed response may make an assignment faster, while the learner spends less time working through the steps. That is the concern behind calls to distinguish useful assistance from reliance that weakens independent thinking.

High school students' reports of weaker thinking add urgency to that discussion. Their responses may reflect awareness that AI has become deeply embedded in their study routines. Educators must consider both the immediate usefulness of the tool and whether students continue developing the skills the assignment is meant to teach.

Medical Students Show a Gap Between Use and Training

A study of first year medical students at Chiba University, published in a scientific journal, provides a view of AI use after secondary school. The survey, conducted in April 2025, found that 84.7 percent used generative AI. Language learning was a leading use at 44.9 percent, followed by information searches at 39.0 percent.

Despite widespread adoption, the medical students held neutral attitudes toward using AI for assignments. Their responses suggest a more cautious approach to academic integrity than the complete reliance reported among some secondary school students. Academic integrity involves completing work honestly and following the rules for assistance.

Only 49.2 percent of the medical students had received any formal education about AI. Students mainly learned through browsing or instruction from peers, showing how readily practical use can spread without organised teaching.

Students who had received structured instruction showed greater awareness of possible bias in AI outputs and more developed skills in critically evaluating answers. Bias means that a response may reflect an uneven or prejudiced view rather than treating information fairly.

The Chiba findings support the case for deliberate AI education. Familiarity with an interface and frequent use are different from learning how to check information, identify limitations and decide when assistance is appropriate. The study offers a model for secondary education that combines access with careful evaluation.

Universities Move Beyond Blanket Restrictions

Japanese universities have increasingly acknowledged generative AI in their course documents. In 2023, researchers examined approximately 1.35 million syllabi from 707 universities and found just 528 mentions. Most of the courses mentioning the technology restricted or prohibited its use.

By 2024, researchers found 7,442 references in 1.24 million syllabi from 679 universities. The number of mentions rose about fourteenfold, even though the later study examined fewer syllabi and fewer institutions. A syllabus sets out a course's content and expectations, so references to AI show how teaching rules and plans are adapting.

Universities also broadened their approaches. Some courses continued to prohibit AI, while others treated it as a collaborative partner or taught students to evaluate the information it produces. The shift reflects growing recognition that institutions need practical ways to teach responsible use as the technology becomes harder to exclude.

Secondary school rules remained uneven. Just under 10 percent of students in the Kankō Gakuseifuku survey said they did not use generative AI for schoolwork or could not use it because their school prohibited it.

Different rules create different educational experiences. Some students practise navigating research with AI assistance, while others face strict restrictions. That variation may affect their preparation for universities and professional environments where competence with AI is expected.

Teaching Students to Stay in Charge

Gakken's adoption figures, Kankō Gakuseifuku's findings on dependence and Chiba University's findings on training point toward a shared educational need: students require more than permission to use AI or an instruction to avoid it. They need practical competence, critical evaluation skills and ethical reasoning.

Schools can address those needs by teaching students to verify information, recognise bias and use AI to support their own reasoning. The Chiba study's association between structured instruction and more careful evaluation gives that approach a concrete basis. Its model places critical thinking alongside access to the technology.

Japan's experience offers both a warning and a guide for other countries navigating the same change. Widespread use shows the appeal of AI even as institutional rules vary. Reports of weaker thinking among high school students underline the concern that unguided use may reduce the practice needed to develop independent abilities.

Educators therefore face a question of how to keep students active in learning. AI can help with explanations, research and creative work, but the learner still needs to judge the response and take responsibility for the task. Structured teaching seeks to preserve that role as AI becomes an ordinary part of education.

At a Glance

• Gakken's November 2025 survey found conversational AI use among 73.7 percent of high school students, 43.2 percent of junior high students and 36.6 percent of elementary pupils.
• Studying and homework were the leading purpose among high school users at 42.3 percent; 26.0 percent primarily sought information.
• Elementary pupils also used AI creatively, with 23.7 percent generating illustrations or images.
• Kankō Gakuseifuku found nearly 80 percent of teenagers used AI frequently or occasionally, while approximately 20 percent relied entirely on it for answers or calculations.
• Consultation or conversation was reported by 49.9 percent of girls and 23.0 percent of boys.
• No particular change in thinking was the most common response at every school level; high school students reported weaker thinking more often than stronger thinking.
• At Chiba University, 84.7 percent of first year medical students used generative AI, but only 49.2 percent had received formal AI education.
• University syllabus references rose from 528 in 2023 to 7,442 in 2024, alongside a wider range of approaches to teaching and regulating AI.