From brainstorming to plagiarism anxiety — a global survey of 23,218 university students across 109 countries reveals a generation navigating AI with curiosity, hope, and real unease.

Chart 1 — Adoption is high but uneven across the globe

Nearly three in four students surveyed (71.5%) have used ChatGPT at least once. But adoption rates vary widely by region — Latin American students lead at 78%, while Middle Eastern and African students trail at 66%. North America’s low count reflects very few survey respondents, not lower adoption.

Source: Ravšelj et al. (2024). Global ChatGPT Student Survey. Mendeley Data. Q4 (country of study), Q13 (ever used ChatGPT). North America excluded from trend comparisons due to small sample (n = 54).

Chart 2 — What students actually use it for depends on what they study

Brainstorming is the top use case across all disciplines — but beyond that, usage patterns diverge. Applied Sciences students are the heaviest users of coding assistance and study help, while Arts & Humanities students lean on it for proofreading and creative tasks. Hover over any cell to explore the detail.

Source: Ravšelj et al. (2024). Q10 (field of study), Q18a–l (ChatGPT task frequency). ChatGPT users only (n = 16,010). Scale: 1 = Never, 5 = Always.

Chart 3 — The emotional reality: curiosity leading, anxiety following

When students use ChatGPT, they most often feel curious (mean 3.43/5), calm (3.27), and happy (3.12). But negative emotions are present too — confusion (2.54) and pride’s absence suggest students aren’t fully confident in their AI-assisted work. Toggle between positive and negative emotions using the legend to explore the split.

Source: Ravšelj et al. (2024). Q32a–o (emotions while using ChatGPT). ChatGPT users only (n = 16,010). Scale: 1 = Never, 5 = Always.

Chart 4 — Students see real benefits, but fear real risks

This is the central tension of AI in education. Students strongly believe ChatGPT improves their access to knowledge and overall learning — yet they simultaneously fear it will encourage plagiarism, spread misinformation, and hinder deep learning. Both scales run 1–5. Use the dropdown to filter by field of study and see how perceptions shift across disciplines.

Source: Ravšelj et al. (2024). Q26a–j (learning outcomes, users only, n = 16,010); Q22a–j (ethical concerns, all respondents, n = 22,414). Scale: 1 = Strongly disagree, 5 = Strongly agree.

Chart 5 — Students expect disruption, but not doom

On the labour market, students are pragmatic rather than panicked. They strongly agree ChatGPT will change jobs, demand new AI skills, and boost productivity — but they’re less convinced it will simply replace jobs. The chart below plots each belief by its average agreement score (1–5) and colours it by the nature of the predicted change. Click any category in the legend to isolate it.

Source: Ravšelj et al. (2024). Q30a–l (labour market beliefs, all respondents, n ≈ 12,500–13,500 per item). Scale: 1 = Strongly disagree, 5 = Strongly agree.

Data: Ravšelj, D., et al. (2024). Higher Education Students’ Early Perceptions of ChatGPT — Global Survey Data [Data set]. Mendeley Data. https://doi.org/10.17632/gwhdkj5y5k.2
Survey conducted: October 2023 – February 2024 | n = 23,218 students | 109 countries and territories
Visualisations built with R using ggplot2, plotly, dplyr, tidyr, readxl.