1. Introduction

Generative AI tools such as ChatGPT, Gemini, and Claude are becoming widely used in higher education. These tools help students search for information, generate ideas, and complete learning tasks. At the same time, concerns have been raised about overdependence on AI and the possible decline of independent learning skills.

In Vietnam, generative AI is being adopted rapidly by university students. However, there is still limited evidence on whether AI contributes to learning competence and academic motivation. This issue is particularly relevant for Thuy Loi University

Self-Determination Theory (SDT) provides an appropriate framework for this study. According to SDT, motivation develops when three basic psychological needs are satisfied: autonomy, competence, and relatedness. Students are more likely to maintain intrinsic motivation when these needs are fulfilled.

In the Vietnamese setting, comprehensive evidence remains scarce on several fronts: how intensively undergraduates actually use AI; whether such use satisfies the three basic needs; how AI relates to different forms of motivation; and which student groups are most at risk of demotivation. These gaps define the space this study addresses.

Accordingly, this study has four objectives. First, it describes patterns of AI use among students. Second, it examines whether AI use is associated with learning competence. Third, it investigates the relationship between prompting proficiency and motivation. Finally, it identifies student groups that may require additional support.

2. Methods

2.1. Research design

This study used a quantitative cross-sectional survey design. Data were collected through an online questionnaire administered to undergraduate students at Thuy Loi University.

2.2. Participants

The final sample included 295 students. Invalid responses were removed before analysis. Participants were full-time undergraduate students who had experience using at least one generative AI tool. Participation was voluntary.

The sample was predominantly female (86.4 %, n = 255; male 13.6 %, n = 40). By year of study, participants were distributed across Year 1 (51.5 %), Year 2 (39.7 %), Year 3 (7.1 %), and Year 4 (1.7 %). By academic standing (GPA), the distribution was good (56.6 %), very good (19.7 %), average (19.7 %), and excellent (4.1 %).

2.3. Measures

The questionnaire consisted of three sections. The first collected demographic information and AI use. The second measured autonomy, competence, and relatedness using five-point Likert scales. The third assessed academic motivation, including intrinsic motivation, identified regulation, external regulation, and amotivation. All scales were adapted from previously validated instruments.

2.4. Data analysis

Data were analysed in SPSS 26.0. The analyses comprised descriptive statistics (means, standard deviations, and percentages); one-way ANOVA to compare groups defined by frequency of AI use, confidence level, GPA, and year of study; Pearson correlations to test relationships among continuous variables; and independent-samples t-tests to compare male and female students. Statistical significance was set at α =.05.

2.5. Research ethics

The study followed established ethical standards. All participants were informed of its purpose and gave voluntary consent before completing the questionnaire. Responses were collected and processed anonymously.

3. Results

3.1. Sample characteristics and patterns of AI use

Among the 295 respondents, the sample skewed female (86.4 %), reflecting the composition of several programmes at Thuy loi University. AI use was widespread: 73.6 % of students used AI at least three to five times per week, and 26.4 % used it daily, indicating that these tools have become a common feature of everyday study. ChatGPT was the most widely used tool (56.3 %), followed by Gemini (29.2 %). Prompting skill, however, was less developed: only 18.0 % of students rated themselves at Level 3 (proficient), while 53.6 % placed themselves at Level 2 (intermediate) and 28.5 % at Level 1 (basic).

Table 1

Descriptive statistics of the sample (N = 295)

Source: Author's analysis of survey data.

3.2. Mean scores of the study constructs

Table 2 reports the mean score for each construct. The three SDT needs all fell within the “agree” range: autonomy (M = 3.55, SD = 0.73), competence (M = 3.69, SD = 0.78), and relatedness (M = 3.71, SD = 0.78). Motivation scores were likewise fairly positive, with identified regulation highest (M = 3.85, SD = 0.92). Even so, amotivation averaged M = 3.20 (SD = 0.97), placing it in the “uncertain” range and marking it as a concern that warrants attention.

Table 2

Means and standard deviations of the study constructs

3.3. Frequency of AI use, competence, and relatedness

One-way ANOVA showed that frequency of AI use had a statistically significant effect on competence (F = 3.412, p =.018) and relatedness (F = 3.746, p =.012), but not on autonomy (F = 1.038, p =.376) or intrinsic motivation (F = 2.047, p =.107). Students who used AI daily scored 0.35 points higher on competence (M = 3.80) than those who rarely used it (M = 3.46), and 0.41 points higher on relatedness (3.93 vs. 3.52; see Table 3).

Table 3

Construct means by frequency of AI use

Note: * p <.05.

3.4. The role of prompting skill

Among the findings, the influence of prompting skill was the most pronounced. Self-rated confidence in using AI (that is, prompting skill) had a significant effect on all four constructs: autonomy (F = 5.663, p =.004), competence (F = 5.542, p =.004), relatedness (F = 3.068, p =.048), and external regulation (F = 5.793, p =.003). Compared with the Level 1 (basic) group, the Level 3 (proficient) group scored 0.38 points higher on autonomy (3.71 vs. 3.33) and 0.42 points higher on competence (3.89 vs. 3.47; see Table 4). In this study, prompting was the only variable significantly linked to autonomy, whereas frequency of use alone (Section 3.3) showed no such relationship.

Table 4

Construct means by AI confidence level (prompting)

Note: * p <.05; ** p <.01.

3.5. Academic standing and the risk of amotivation

A Pearson correlation revealed a significant negative relationship between GPA and amotivation (r = −.142, p =.014). Very good students recorded the lowest amotivation (M = 2.88), while average-achieving students recorded the highest (M = 3.41) — leaving the average group at nearly twice the risk of the very good group. ANOVA by GPA likewise showed a significant difference in amotivation (F = 3.219, p =.023), with no significant differences on the remaining constructs.

Table 5

Amotivation and competence by academic standing (GPA)

Note: Correlation between amotivation and GPA: r = −.142; p =.014.

3.6. Subgroup differences: year of study, sex, and AI tool

To examine whether motivational profiles varied across the sample, mean construct scores were compared by year of study, by sex, and by the AI tool students reported using most often. The results of all three comparisons are summarised in Table 6.

Table 6

Construct means by year of study, sex, and primary AI tool

Note. Values are construct means on a 5-point scale. A dash (–) indicates that the construct was not among those compared for that grouping variable. The Year 4 subgroup (n = 5) is reported for completeness but is too small to support inference. Sex differences are expressed as male minus female.

Differences by year of study

Analysis by year of study revealed a notable curvilinear pattern. Second-year students scored markedly lower on competence (M = 3.47) and relatedness (M = 3.50) than first-year students (competence M = 3.84; relatedness M = 3.89), a decline of roughly 9–10 % on each indicator. Both indicators recovered in the third year (competence M = 3.88; relatedness M = 3.77), returning to approximately first-year levels. Amotivation followed the same downward movement in Year 2 (M = 3.06, compared with M = 3.33 in Year 1). The apparent further decline in Year 4 should be interpreted with caution, as only five students were in this subgroup.

Differences by sex

Independent-samples t-tests indicated that female students scored higher than male students on external regulation (M = 3.63 vs. 3.40) and on amotivation (M = 3.22 vs. 3.06), whereas male students reported somewhat higher competence (M = 3.84 vs. 3.66). This pattern suggests that female students may experience greater pressure from external sources such as family, teachers, and society when using AI tools for learning.

Comparison across AI tools

A comparison of the two most commonly used tools showed that ChatGPT users scored higher than Gemini users on competence (M = 3.74 vs. 3.57), relatedness (M = 3.80 vs. 3.57), and intrinsic motivation (M = 3.83 vs. 3.65), with gaps of roughly 5–6 % on each measure. Taken together with the patterns above, the results indicate that motivational experiences of AI-supported learning differ systematically by academic stage, by sex, and by the tool students rely on most.

4. Conclusion

Overall, the findings suggest that generative AI is associated with learning competence and academic motivation. Frequent AI use was related to competence and relatedness, but not to autonomy. In contrast, prompting proficiency was associated with all three psychological needs.

The findings have several practical implications. Universities should provide structured training in prompting skills from the first year. Additional academic support may benefit second-year students and students with average academic performance. Universities should also encourage students to use AI critically so that it supports rather than replaces independent learning.

This study has several limitations. First, the cross-sectional design does not allow causal conclusions. Second, the sample included many more female than male students. Third, very few participants were in Years 3 and 4. Finally, all data were self-reported. Future studies should use longitudinal designs and include intervention-based approaches to examine the effects of AI-supported learning more comprehensively.

References:

1. AbuSeileek, A. F., Alotaibi, F. S., Alqadi, M., & Al-Rawashdeh, A. (2024). Higher education students' task motivation in the generative artificial intelligence context: The case of ChatGPT. Information, 15(1), Article 33. https://doi.org/10.3390/info15010033
2. Chiu, T. K. F. (2024). A classification tool to foster self-regulated learning with generative artificial intelligence by applying self-determination theory: A case of ChatGPT. Educational Technology Research and Development, 72(4), 2401–2416. https://doi.org/10.1007/s11423–024–10366-w
3. Deci, E. L., & Ryan, R. M. (2000). The «what» and «why» of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11(4), 227–268. https://doi.org/10.1207/S15327965PLI1104_01
4. Du, J., & Alm, A. (2024). The impact of ChatGPT on English for Academic Purposes students' language learning experience: A self-determination theory perspective. Education Sciences, 14(7), 726. https://doi.org/10.3390/educsci14070726
5. Le Cong Minh, Nguyen Thi My Dung, Nguyen Thi Hong Nhung, & Nguyen Thi Kim Oanh. (2024). The impact of students' using generative AI for learning on self-learning motivation: A study based on self-determination theory. In Proceedings of the International Conference on Research and Development in Education (pp. 201–214). Atlantis Press. https://doi.org/10.2991/978–94–6463–583–6_19
6. Mohamed, A. M., Shaaban, T. S., Bakry, S. H., Ahmed, H. M., & Hassan, A. A. (2025). Empowering faculty of education students: Applying artificial intelligence's potential for motivating and enhancing learning. Innovative Higher Education. https://doi.org/10.1007/s10755–024–09747-z
7. Ng, D. T. K., Tan, C. W., & Leung, J. K. L. (2024). Empowering student self-regulated learning and science education through ChatGPT: A pioneering pilot study. British Journal of Educational Technology. https://doi.org/10.1111/bjet.13454
8. Ryan, R. M., & Deci, E. L. (2020). Intrinsic and extrinsic motivation from a self-determination theory perspective: Definitions, theory, practices, and future directions. Contemporary Educational Psychology, 61, Article 101860. https://doi.org/10.1016/j.cedpsych.2020.101860
9. Nguyen, T. T. T. (2023). Self-determination theory (SDT): Conceptions, classification, and implications for motivating Vietnamese students. Vietnam Journal of Education, 23(10), 33–38. [in Vietnamese]