AI helps students with work, but more use doesn't mean better scores: Study

Artificial intelligence (AI) chatbots are now widely used by 15-year-old students for their schoolwork, according to a recent study. They use the chatbots to learn, research new topics, draft assignments and summarise texts they have been asked to read.

 

The findings were published in the Organisation for Economic Co-operation and Development’s (OECD) Programme for International Student Assessment (PISA) 2025 Results, Volume I: Future-Ready Students, earlier last week.

 

The report examined how students use AI for schoolwork and how different patterns of use relate to science performance.

What is the PISA 2025 study?

According to the report, more than 760,000 students, representing about 33 million 15-year-olds across 91 countries and economies, participated in PISA 2025. The assessment focused on science, along with reading and mathematics. It also assessed learning in the digital world.

  

The PISA target population covered students aged between 15 years and three months and 16 years and two months who are enrolled in school and have completed at least six years of formal schooling.

 

Notably, India is not covered in the PISA 2025 results.

AI use is widespread among students

The report said the rapid development of AI tools has created new opportunities and challenges for education systems.

 

“The rapid development of artificial intelligence (AI) tools, such as chatbots, has created new opportunities and new challenges for education systems. AI chatbots differ from other digital resources analysed in this chapter in that they became available to the public only in late 2022 and their use has expanded across regions and languages since then,” the report said.

 

By 2025, AI chatbots such as ChatGPT were being used for schoolwork by a majority of students in most participating countries and economies. The frequency of use varied between students and countries, and regular use remained limited in some systems.

 

On average across OECD countries, “to help me learn” was the most commonly reported purpose for using AI. It was followed by conducting preliminary research on a new topic, drafting text for written assignments and summarising a text that students had to read.

 

The report also said almost-daily use remained relatively uncommon amongst the students. About 14 per cent of students across OECD countries said they never or rarely used AI for any of the schoolwork-related purposes examined by PISA.

 

In about 20 countries of 91 who were covered in the assessment, a majority of students said they used AI to summarise assigned reading texts only once or twice a year or less, or never. The same was true for drafting written assignments in 20 countries and economies.

Students use AI for reading, research and writing

PISA collected data on four specific uses of AI in schoolwork: summarising assigned reading, conducting preliminary research, drafting written assignments and using AI “to help me learn”.

 

The report said students who do not use AI for schoolwork generally score higher in science than students who do. They are also more likely to report higher levels of help-seeking behaviour.

 

However, the relationship changes when AI is used specifically to help students learn. Students who use AI for this purpose on a weekly basis tend to have similar science performance to non-users after accounting for socioeconomic background. Students who use it almost every day or more, as well as those who use it twice a month or less, tend to have lower scores than non-users.

 

For specific tasks, non-users generally outperform users. Among students who use AI to summarise texts or conduct preliminary research, those using it once a month to twice a week outperform students who use it twice a year or less and those who use it almost every day or more. For drafting written assignments, the advantage for moderate users is less pronounced.

 

The report said these results do not establish that AI use causes lower science performance.

 

“These relationships do not necessarily imply a negative impact of AI use on science performance, but may reflect a complex mix of who adopts AI and how they use it,” it said.

 

The report also said that the findings suggest that moderate and intentional use of AI for schoolwork and learning may be positively related to student performance.

AI use and curiosity

PISA also found an association between AI use and students’ reported curiosity.

 

Students who said they used AI every day or almost every day, regardless of the task, also reported higher levels of curiosity on average across OECD countries. Students with more limited AI use reported lower levels of curiosity, although the relationship between curiosity and not using AI was less consistent.

 

The OECD said existing research on curiosity and AI use among students remains limited. It says further research is needed to determine whether AI provides more opportunities for personalised inquiry to students who are already highly curious or whether AI can help develop curiosity.

AI use varies across education systems

The frequency of AI use also varies among education systems, including those that perform above the OECD average in science.

 

The report said differences in AI use among higher-performing systems may partly be linked to their policies on digital resources and AI.

 

“Korea and Singapore have large-scale policies supporting digital learning, with Singapore explicitly including AI tools. Japan and Estonia, where AI is used less frequently for schoolwork, have less comprehensive digital-learning policies,” the report said.

 

The report also said lower-performing systems show less variation in AI use than higher-performing systems. It gave as a possible hypothesis that some lower-performing systems may have adopted AI more quickly as a compensatory strategy.