A study conducted on 26,000 students across 12 countries shows that ChatGPT can improve academic performance, but only under the condition of moderate use. Overusing AI leads to a decline in motivation, plagiarism, and problems with critical thinking. How are schools and universities responding to this issue in 2026?
In 2026, artificial intelligence has become an inseparable element of education – from doing homework to preparing theses. ChatGPT and similar tools have revolutionized the way students acquire knowledge, but their impact on the learning process is raising more and more controversy. A recent study published in Scientific Reports sheds new light on this issue, showing both the benefits and serious risks associated with the abuse of AI in schools.
Study of 26,000 students: What is really happening in classrooms?
An international team of researchers from the University of Oxford, the University of California, Berkeley, and the University of Warsaw analyzed data collected from 26,245 students aged 12–19 from 12 countries, including Poland. The study was conducted between September 2024 and December 2025, and its results were published in March 2026. The goal was to examine how frequent use of ChatGPT affects academic performance, critical thinking skills, and motivation for independent work.
Short-term benefits vs. long-term risks
The study shows that students using ChatGPT 1–2 times a week achieved 15–20% better results in tests for STEM subjects, such as mathematics or physics. The tool helped them quickly explain difficult concepts and fill gaps in their knowledge. However, more frequent use of AI had the opposite effect.
Students who reached for ChatGPT daily or several times a day had 30% lower scores on tasks requiring critical thinking, such as text analysis or argumentation. Worse, 42% of respondents admitted that they less frequently look for solutions on their own if they can use AI assistance. This phenomenon raises concerns about the long-term consequences for the cognitive skills of young people.
Plagiarism and superficial knowledge
One of the most serious problems identified in the study is superficial knowledge acquisition. As many as 28% of students admitted to copying answers from ChatGPT without understanding their content, and 60% did not verify the accuracy of the information provided by the tool. This phenomenon is particularly dangerous in the context of humanities and social sciences, where the ability to critically analyze sources is key.
In Poland, this problem is even more visible. The study involved 3,200 Polish students, of whom 38% used ChatGPT daily – which is 13 percentage points higher than the global average. Polish students demonstrated significantly worse essay writing skills, which may have long-term consequences for the quality of education in the country.
Educational inequalities exacerbated by AI
The study also revealed that AI can exacerbate educational inequalities. Students with poorer access to technology, for example from small towns or families with lower socioeconomic status, had difficulty catching up when using AI became the standard in education. In developing countries, this problem is even more acute because paid versions of AI tools, such as ChatGPT Plus, are inaccessible to many students.
The stance of ChatGPT creators: Help, but with limitations
OpenAI, the company responsible for the development of ChatGPT, has long emphasized that their tool is intended to serve as support, not a replacement for independent learning. In April 2025, an official report titled "AI in Education: Opportunities and Challenges" was published, presenting guidelines for the ethical use of AI in schools.
"Educator" mode and collaboration with UNESCO
In response to growing concerns regarding the abuse of ChatGPT, OpenAI introduced the "Educator" mode in 2025, which limits the tool's functions. For example, in this mode, ChatGPT does not provide ready-made answers to exam questions, but instead asks guiding questions that are meant to help the student find the solution on their own.
The company is also collaborating with UNESCO to develop global guidelines for the use of AI in education. The goal is to create a framework that will allow for the safe and effective use of AI tools in schools around the world.
Statements from OpenAI representatives
Sam Altman, CEO of OpenAI, during the SXSW 2026 conference in March of this year, emphasized that AI will not replace teachers, but can relieve them of routine tasks. In his view, it is crucial to teach young people how to use these tools wisely so that they do not become a substitute for independent thinking.
A similar stance is presented by other companies developing AI models, such as Google (Gemini) and Microsoft (Copilot). Both introduced educational tools in 2025 with AI plagiarism detection features and a "Learning Coach" mode, which helps students by asking questions instead of providing ready-made answers.
Scale of AI usage in education: Data from Poland and the world
In 2026, the use of AI in education became widespread, but the scale of this phenomenon varies depending on the region. According to the UNESCO report "AI in Education: Global Trends 2026", 68% of students in developed countries use AI tools at least once a week. In East Asia, this percentage is even higher – in South Korea, it reaches 82%, and in China 78%.
Poland: At the forefront of AI abuse
In Poland, the situation is particularly worrying. According to a report by the Ministry of Education and Science from March 2026, 52% of students use ChatGPT at least once a week, and 20% of schools have introduced bans or restrictions regarding the use of AI in homework. The problem, however, is the lack of uniform guidelines at the national level – decisions on this matter are made at the level of individual institutions.
It is also worth noting that only 15% of teachers in Poland feel competent in working with AI tools. This shows how important it is to train teaching staff in the field of new technologies so that they can effectively supervise the use of AI by students.
USA: AI dominates in high schools
In the United States, the situation is similar. According to a study by the National Education Association (NEA) from January 2026, 75% of high school students use AI to do their homework. At the same time, 40% of teachers admit that they do not feel prepared to work with students using these tools. This shows the significant gap that exists between the pace of AI implementation in education and the preparation of teaching staff.
Risks associated with ChatGPT abuse
Experts identify several key risks associated with the abuse of ChatGPT and similar tools in education. Here are the most serious ones:
1. Decline in critical thinking
A study conducted by the Harvard Graduate School of Education in 2025 showed that students using AI had 25% worse results on tasks requiring source analysis and critical thinking. This phenomenon is particularly dangerous in the context of humanities, where the ability to interpret texts and formulate one's own opinions is key.
2. Plagiarism and superficial knowledge
The anti-plagiarism tool Turnitin reports that in 2026, 1 in 3 student papers contains AI-generated fragments. Worse, many students do not verify the accuracy of the information provided by AI tools, which leads to superficial knowledge acquisition and factual errors.
3. Technology dependency
A University of Cambridge study from 2025 revealed that 18% of students admitted they cannot write an essay without AI help. This phenomenon is particularly worrying because it shows how much young people are becoming dependent on technology, losing the ability to think and work independently.
4. Educational inequalities
AI can exacerbate educational inequalities because students from poor families often do not have access to paid versions of AI tools, such as ChatGPT Plus. According to an OECD report from 2026, differences in access to technology can lead to widening gaps in educational outcomes between students from different backgrounds.
5. Misinformation
A Stanford Internet Observatory study from 2025 found that 22% of ChatGPT's answers to questions in the fields of history and social sciences contained errors or biased information. This shows how important it is to teach students a critical approach to information generated by AI.
How are schools and universities responding to the problem?
Faced with growing concerns regarding AI abuse, many schools and universities around the world are introducing regulations regarding the use of these tools. Here are a few examples:
Poland: Bans and restrictions
- The University of Warsaw introduced a ban on the use of AI in theses in 2025 without the supervisor's consent. The university uses the Originality.AI tool to detect AI-generated texts.
- Kopernik High School in Warsaw has required since 2026 that all homework must be submitted in written form, without the use of a computer.
USA: Blocks and guidelines
- New York City Public Schools blocked access to ChatGPT on the school network in 2025, but in 2026 they partially unblocked it under teacher supervision.
- University of California, Berkeley requires disclosure of AI use in academic work.
France: National teacher training program
The French Ministry of Education introduced a national teacher training program in 2026 regarding AI and anti-plagiarism guidelines. The goal is to prepare teaching staff to work with students using AI tools.
China: Ban on AI use in primary schools
The Chinese Ministry of Education introduced a ban on the use of AI in primary schools in 2025 to protect younger students from the negative effects of abusing these tools.
Alternative tools and educational methods
Although ChatGPT and similar tools can be helpful, there are also alternative methods and technologies that support the learning process without the risk of AI dependency. Here are a few of them:
AI tools supporting critical thinking
- Khanmigo (Khan Academy) – instead of giving ready-made answers, it asks guiding questions that help students find the solution on their own. More about Khanmigo.
- Socratic by Google – helps in solving tasks step-by-step, explaining the thought process. More about Socratic.
Methods without AI
- Flipped Classroom – students learn theory at home, and in class, they solve tasks under the teacher's supervision.
- Project-Based Learning (PBL) – students work on long-term projects, developing creativity and teamwork skills.
- Debates and discussions – developing argumentation and critical thinking skills.
Educational initiatives
- Poland: "Digital Student" Project – training for teachers in the field of AI. More about the project.
The future of AI in education: What awaits us by 2030?
The prospects for AI development in education for the coming years are both promising and worrying. Experts predict several key trends:
Personalization of learning
AI will adapt the pace and methods of teaching to the individual needs of students. A McKinsey report from 2026 predicts that by 2030, personalized learning paths will become the standard in developed countries. This could significantly improve learning efficiency, but it will require appropriate teacher preparation.
Virtual teachers
Experiments with AI avatars conducting lessons are already underway. Projects such as Synthesia or Duolingo Max show what the future of education might look like. However, experts emphasize that virtual teachers will never replace human interaction, which is crucial for the social and emotional development of students.
Ethics and regulations
The European Union plans to introduce legislation regarding AI in education by 2028 as part of the AI Act 2.0. The goal is to create a framework that will allow for the safe and ethical use of AI in schools. Similar regulations are being developed in other regions of the world.
Democratization of access
AI tools will become free and widely available, but their effective use will require appropriate training for teachers and students. A World Economic Forum report from 2026 emphasizes that it will be crucial to prevent the widening of educational inequalities related to access to technology.
Summary: AI in education – opportunity or threat?
The study of 26,000 students clearly shows that ChatGPT and similar tools can be a valuable support in learning, but only under the condition of moderate and conscious use. Overusing AI leads to a decline in motivation, plagiarism, and problems with critical thinking, which may have long-term consequences for the quality of education.
It is crucial to prepare teachers and students to work with new technologies. Schools and universities must introduce clear regulations regarding the use of AI, as well as promote alternative teaching methods that develop independence and creativity. Only in this way will it be possible to harness the potential of AI while minimizing the risks associated with it.
In the context of the future of education, AI will certainly play an important role, but it will not replace the human factor. How we use its capabilities depends on us – teachers, students, and policymakers.
If you are interested in the topic of AI's impact on other aspects of life, check out our previous posts on artificial intelligence and the labor market as well as the future of university degrees in the era of algorithms.
Comments