15 Sept 2026 /
Insights
Mini report · STEM students in Malaysia and Singapore
Two neighbours, one policy gap, and the same collapse in confidence.
An excerpt of the Studiosity Global Student Wellbeing Survey 2026, conducted with YouGov. Malaysian and Singaporean STEM students use AI at identical rates and trust their own learning equally little, but one group has been given roughly twice as much institutional guidance as the other. Only findings that are notably different from other STEM cohorts, from students in other fields, or from each other are reported here.
How to read this report. Every figure states its group. Malaysian STEM means the 210 Malaysian respondents studying STEM subjects. Singaporean STEM means the 208 Singaporean respondents studying STEM subjects. Combined means both together, 418 students. Global STEM means all 2,694 STEM students across the nine markets surveyed.
Malaysian STEM
210 students, 204 AI-users
210 students, 204 AI-users
Singaporean STEM
208 students, 202 AI-users
208 students, 202 AI-users
Combined
418 students, 406 AI users
418 students, 406 AI users
Ranked AI stresses
317 of the combined group gave their rankings
317 of the combined group gave their rankings
Combined
97%
have used AI for study — 97% in each market separately
Global STEM 88%
Combined
33%
are confident they are learning — Malaysia 33%, Singapore 34%
Global STEM 45%
The divergence
36/66
per cent given AI guidance by their uni: Malaysian STEM 36%, Singaporean STEM 66%
Global STEM 60%
Combined
50%
are moderately or worse stressed about being wrongly flagged by AI detection tools: Malaysia 51%, Singapore 50%
Global STEM 42%
01 The combined picture
The most AI-native STEM students in the survey are the least sure they are learning.
Ninety-seven per cent of both Malaysian and Singaporean STEM students have used AI for assignments or study tasks, against 88% of STEM students globally. Combined, 41% use it regularly, against 33% globally, though the two markets differ here: 48% of Malaysian STEM students use AI regularly, against 34% of Singaporean STEM students.
Confidence is where the two markets converge most sharply, and where both sit lowest in the study. A third (33%) of Malaysian STEM students and 34% of Singaporean STEM students are very or extremely confident they are learning while using generative AI. No other STEM cohort in the survey falls below 41%; the United States reaches 51% and Saudi Arabia 61%.
Combined, 78% report stress from using AI as part of their studies, against 64% of STEM students globally, and 88% report at least some stress about being wrongly flagged for misconduct by detection tools, against 77% globally. Both markets are near the top of the research on these measures.
Combined, 49% agree they would use AI to write their papers if there were no rules against it, against 37% of STEM students globally. Malaysian STEM 51%, Singaporean STEM 47%.
“AI is easy to use but the real world is not. If I get used to AI, the world will eat me up alive.”
Singapore, Central Region · 18–25 · 3rd year · STEMOn AI and critical thinking
“I feel I offload too much of my mental load there.”
Malaysia, Selangor · 26–33 · Postgraduate · STEMOn AI and critical thinking
02 The natural experiment
Singapore has given nearly twice as many students AI guidance. Confidence and detection stress are unchanged.
Two thirds (66%) of Singaporean STEM students say their institution has provided training, policies, or guidelines on the ethical use of AI, above the global STEM figure of 60%. Thirty-six per cent of Malaysian STEM students say the same, the lowest of any STEM cohort in the surveyed regions, and 30% of them are not sure whether their institution has, against just 15% in Singapore.
That is a gap in institutional guidance around AI, between two neighbouring markets with identical AI adoption. The main measure where this deviates between the cohorts is confidence in the institution's ability to adapt. Seventy-four per cent of Singaporean STEM students say their institution is adapting fast enough to include AI support tools, against 55% of Malaysian STEM students.
This mirrors the global pattern reported in the main study, and it holds within these two markets even though they differ on almost every other variable. Guidance buys visible institutional credibility. It does not, on this evidence, buy a student's confidence in their own learning or relieve their anxiety about detection.
Combined group, split by whether guidance was provided
✅ Guidance provided (212 students)
80% say the institution is adapting fast enough
37% confident they are learning
❌ No guidance (113 students)
46% say it is adapting fast enough
30% confident they are learning
➖ Not sure (93 students)
52% say it is adapting fast enough
25% confident they are learning
Sub-bases of 93 to 212 students. The confidence effect is stronger in Malaysia, where guidance is scarce, at 41% against 30%, than in Singapore, where it is already widespread, at 34% against 30%. Read as directional.
“Only got a form asking us to declare. Doesn't mean they truly accept it if our AI content is high.”
Singapore, West Region · 18–25 · 3rd year · STEMOn why their institution is not adapting fast enough
“Even if our lecturers are mainly open-minded about the usage of AI, the school spent little to no amount on setting black and white guidelines for using AI ethically.”
Malaysia, Sarawak · 18–25 · 1st year · STEM · English not first languageOn why their institution is not adapting fast enough
03 What each cohort fears
Same anxiety, different name: Malaysian students blame themselves, Singaporean students blame the process.
Both groups carry more 'detection stress' than any other STEM cohort in the surveyed regions. What they call it when asked to name their three biggest AI stresses is where they differentiate.
Malaysian STEM students put "becoming addicted to AI" first, with 57%, the highest of any STEM cohort in the study. "Being accused of cheating when they did nothing wrong" comes fifth for them, level with "ownership of their own work", at 33% — the lowest rate of any STEM cohort. Singaporean STEM students invert that order: 'false accusations of cheating' is their top stressor at 48%, "ownership of work" second at 45%, and addiction to AI is third at 41%. The Singapore STEM cohort are also almost twice as likely as the Malaysian group to cite "getting caught cheating", at 23% against 13%.
Stresses of using AI, named in a student's top three
Malaysian STEM
Singaporean
Global STEM
Becoming addicted to AI
57%
41%
44%
Being accused of cheating when I did nothing wrong
33%
48%
44%
Ownership of my own work
33%
45%
38%
Not learning anything
41%
35%
38%
Privacy or security of my data
40%
35%
39%
Other students using AI to get better grades
38%
28%
30%
Being unsure about the AI rules at my university
25%
30%
29%
Getting caught cheating
13%
23%
19%
Bases: 162 Malaysian STEM, 155 Singaporean STEM, and 1,518 global STEM students who use AI and report any stress in using it, each naming the three factors that stress them most from ten options. Bars show the Malaysian column only, on a nought to 60% axis; the Singaporean and global columns are given as figures.
INSIGHT · The blind spot in the ai arms race
The data shows that simply catching students using AI doesn't solve the underlying crisis of authentic learning. When universities rely on a "police and punish" model, they fail to address the two thirds of STEM students who aren't confident they're developing their own skills when using AI.
By moving from AI detection to formative support and confirming evidence of authorship, Studiosity helps students self-validate that they have actually done the thinking. Our platform flips the focus back to learning capacity, ensuring that AI tools serve to "Help me learn" rather than just "Get it done faster".
04 Five interesting insights from the report
1. Both markets want peer mentors, and neither has enough · combined
Fifty-seven per cent of combined students without a mentor say they need one, against 47% of STEM students globally. Singaporean STEM records the higher demand at 60%, Malaysian STEM 54%. Mentor availability itself is unremarkable, at 42% and 40% against 40% globally; it is the unmet demand that deviates.
2. Weaker belonging and weaker access to educators than any comparable cohort · combined
Sixty-one per cent of combined students feel they belong in their university community, against 71% of STEM students globally. Seventy-seven per cent are confident contacting their professor, against 84%. Malaysian STEM 60% and 79%, Singaporean STEM 63% and 75%; Singaporean STEM records the lowest confidence contacting an educator of any STEM cohort in the study.
3. Study stress diverges sharply, AI stress does not · by region
Thirty-nine per cent of Singaporean STEM students feel stressed by studying daily or constantly, against 25% of Malaysian STEM students, the lowest of any STEM cohort in the study. Yet stress from using AI is nearly identical, at 77% and 79%. Whatever is driving Singaporean study pressure is not the same thing that is driving AI pressure in either market.
4. A 45-point gap between the grade and the conviction · Malaysian STEM
Seventy-eight per cent of Malaysian STEM students agree a good grade means they learned the subject well, the highest of any STEM cohort, against 33% who are confident they are actually learning with AI. Singaporean STEM students hold the two beliefs closer together, at 64% and 34%, and their 64% is among the lowest faith in grades recorded in the study.
5. Both expect the labour market to demand AI, and to be reshaped by it · combined
Sixty-five per cent of STEM students agree AI skills will be important for getting a job in their field, against 60% of STEM students globally, and 51% see AI as a threat to jobs in their field, against 46%. Singaporean STEM is higher on both, at 69% and 55%.
“Sometimes the lecture slide are hard to understand due to the usage of advanced languages.”
Malaysia, Sarawak · 18–25 · 2nd year · STEM · English not first languageOn why they mainly use AI
“AI is quite good at explaining concepts that lecturers cannot explain or cannot be bothered to explain.”
Singapore, Central Region · 18–25 · 2nd year · STEMOn why they mainly use AI
“Most of the times there is no need to contact them. Even if I do have something to ask, the fear of asking easy questions via email means that I don't contact them at all anyway.”
Singapore, North-East Region · 26–33 · 3rd year · STEMOn contacting their professor or teacher
“If you just copy and paste the question given to you to AI prompt and ask it to write it for you without even reading what it gave to you, of course it will reduce your critical thinking. AI should be used to assist you with your work, not just blindly pasting the answer.”
Malaysia, Pahang · 18–25 · 3rd year · STEM · English not first languageOn AI and critical thinking
06 Qualitative themes
Malaysian STEM students gave 118 free-text answers on AI and critical thinking, Singaporean STEM students 112. Themes are listed in order of how often they appeared within each market, without percentages, since they were not offered to students as choices.
Malaysian STEM
210 students
1. Ethical or not depending on the user
Responsibility is located in the student, not the tool or the rules. Many volunteer a personal standard for acceptable use unprompted.
2. AI as a translator of academic English
Summarising dense slides, simplifying research papers, explaining terminology. Distinctive to this cohort.
3. Rules set by individual lecturers
Permission, prohibition, and detection thresholds arrive lecturer by lecturer. Several learned the rules from peers or seniors.
4. Dependency described as laziness
Addiction, laziness, slowing down, offloading. Self-criticism rather than complaint, matching the 57% who name 'addiction' as their biggest concern.
5. Asking for mentoring and industry exposure
Mentor-mentee programmes, internships, up-to-date laboratory equipment, clearer lecturer communication. Very few asked for more AI.
Singaporean STEM
208 students
1. Declaration as the operative rule
Students describe forms, compulsory modules, and declaration statements. Several note that declaring is not the same as being trusted.
2. Over-reliance stated plainly
Depending on AI for small things, difficulty starting to write without it, and reduced memory retention. Named directly and without hedging.
3. AI as a substitute for teaching quality
Explaining what a lecture did not. This is a criticism of instruction, expressed as a use case for AI.
4. Policy exists but is not evolving
Where students say the institution is not adapting, the reason given is usually stagnation rather than absence: guidelines were issued and then left alone.
5. Degree as transaction, not community
On belonging, students describe attending for the qualification and the job. Requests for improvement centre on mentoring, industry alignment, and career guidance.
07 Method and questions
Bases. Malaysian STEM 210 students, of whom 204 use AI and 162 reported any stress in using it. Singaporean STEM 208 students, of whom 202 use AI and 155 reported any stress. Combined 418, 406, and 317. Drawn from Malaysian and Singaporean samples of 531 and 541, and a total sample of 10,329.
Malaysian STEM profile. 158 aged 18 to 25; 127 men, 83 women; 80 third year, 51 postgraduate, 42 first year, 37 second year. Selangor 66, Kuala Lumpur 19, Penang 18, Johor 16, Perak 14, other states smaller.
Singaporean STEM profile. 117 aged 18 to 25, 45 aged 26 to 33, 33 aged 34 to 41; 139 men, 65 women; 84 third year, 71 postgraduate, 27 second year, 26 first year. West 49, Central 40, North-East 38, North 35, East 32.
Comparators. Global STEM refers to all 2,694 STEM students across the nine markets surveyed. Country STEM comparators range from 102 students in New Zealand to 671 in the United States.
Rounding. All percentages are rounded to the nearest whole number, so components may not sum to 100. Sub-bases under 100 are stated and treated as indicative. Fieldwork was conducted online by YouGov for Studiosity in December 2025.
Questions used. QA1, QB4, QC1, QC2, QC5, QD1, QD2, QD5, QD7, QD8, QD10, QD11, QD12, QD13, QE1, QF1, and profiling questions QS4, QS7, QS10, QS13, and QS15. Full wording is given in the research index of the global report.
Citation. Studiosity Global Student Wellbeing Survey 2026, conducted with YouGov. Malaysian STEM sub-sample n = 210; Singaporean STEM sub-sample n = 208.
Talk to us about what this means for your students.
Institutional partnerships
Navin - nramalingam@studiosity.com
Navin - nramalingam@studiosity.com
Media and research enquiries
hello@studiosity.com
hello@studiosity.com
Full report
Studiosity Global Student Wellbeing Survey 2026, conducted with YouGov
Studiosity Global Student Wellbeing Survey 2026, conducted with YouGov