Filipino college students find ChatGPT useful and easy to use for academic purposes, but dissatisfaction with subscription fees and an unexpected inverse relationship between ease of use and satisfaction suggest that perceived simplicity may undermine academic rigor in their eyes.
Objective: This study examined Filipino college students' acceptance of and satisfaction with ChatGPT as a learning tool in higher education using the Technology Acceptance Model (TAM). The researchers aimed to understand students' perceived usefulness, perceived ease of use, satisfaction levels, and continuance intention regarding ChatGPT, as well as the correlations among these variables in the Philippine educational context.
Methods: The researchers employed a quantitative, descriptive-correlational, cross-sectional research design. They used convenience sampling to recruit 576 undergraduate students from three higher education institutions in Metro Manila, Philippines: Mapúa University, Central Luzon State University, and Colegio de Muntinlupa. Data collection occurred between May and June 2025 through an online Google Forms survey. The instrument was adapted from Han and Sa (2022) and contextualized for ChatGPT use within the TAM framework. The survey used a four-point Likert scale (strongly disagree to strongly agree) without a neutral option to prevent satisficing behavior. The questionnaire assessed four TAM-related dimensions: perceived usefulness, perceived ease of use, satisfaction, and continuance intention. The instrument demonstrated excellent internal consistency with Cronbach's alpha of 0.950. Participants were required to have prior experience using ChatGPT in academic contexts. The sample was heavily skewed toward engineering students (90%), with respondents primarily aged 18-25 (92%). Data analysis was conducted using jamovi and R statistical software, employing descriptive statistics for Research Question 1 and Pearson's r correlation coefficients for Research Question 2. The study received ethical clearance from Central Luzon State University (ERC Code 2025-625F) and adhered to the Philippine Data Privacy Act of 2012.
Key Findings: Students demonstrated high acceptance of ChatGPT across most dimensions. Perceived ease of use scored highest (M=3.03, SD=0.66), with students reporting they clearly understood how to use ChatGPT, could operate it skillfully, found learning it easy, and considered it easy to use overall. Perceived usefulness also scored highly (M=2.96, SD=0.67), with students affirming that ChatGPT efficiently provides useful and interesting educational content and helps improve educational outcomes. Continuance intention was moderate (M=2.80, SD=0.69), indicating students intend to use ChatGPT in the future, will use it again, plan to speak positively about it, and will recommend it to others. However, satisfaction levels were notably lower (M=2.68, SD=0.70), primarily due to strong disapproval of subscription fees (M=2.37). While students were satisfied with choosing, using, and ChatGPT overall, the cost barrier significantly reduced overall satisfaction.
The correlation analysis revealed unexpected patterns that challenge traditional TAM assumptions. A strong positive correlation existed between satisfaction and perceived usefulness (r=0.915, p=0.085), and moderate positive correlations were found between continuance intention and perceived usefulness (r=0.673, p=0.327), as well as between continuance intention and satisfaction (r=0.693, p=0.307). Most surprisingly, inverse relationships emerged between perceived ease of use and both satisfaction (r=−0.796, p=0.204) and continuance intention (r=−0.588, p=0.412). This contradicts traditional TAM predictions where ease of use typically correlates positively with satisfaction and intention. Despite some strong correlations, none reached statistical significance. The researchers suggest three possible explanations for the inverse ease-of-use relationship: students may perceive excessive ease as trivializing their learning process, cultural learning expectations may equate difficulty with meaningful learning, or concerns about authenticity and academic integrity may arise when tools feel "too easy."
Implications: The findings have important theoretical and practical implications for AI integration in Philippine higher education. Theoretically, the study extends TAM by revealing contextual nuances specific to AI adoption in developing educational contexts, challenging Western-centric assumptions about technology acceptance. The inverse relationship between perceived ease of use and satisfaction suggests that usefulness alone does not guarantee adoption unless core user concerns are addressed. This indicates that emotional and cognitive factors—particularly trust and perceived risk—play more critical roles than previously understood in the TAM framework. The findings support evolving theoretical discussions on AI adoption dynamics where cultural and economic factors intersect with technology acceptance.
Practically, the results underscore the urgent need for educational practitioners and policymakers to develop critical frameworks for AI integration that address broader social implications, potential biases, and power dynamics. Rather than viewing ChatGPT as a standalone tool, stakeholders should recognize it as embedded in the co-evolution of societal structures and educational practices. The study highlights the need for assessment tools specifically designed to understand ChatGPT's impact within the Philippine education system, frameworks that harness AI benefits while mitigating risks to academic integrity, and strategies to address economic barriers (subscription fees) that limit equitable access. Institutions must balance leveraging AI's scalability and immediacy while retaining human feedback's empathetic, contextual, and nuanced qualities through hybrid approaches.
Limitations: The study acknowledges several significant limitations. First, convenience sampling in Metro Manila may not represent students across the Philippine archipelago, particularly those in areas with limited technological access. The sample was heavily skewed toward engineering students (90%), limiting generalizability across disciplines. Second, the cross-sectional design provides only a snapshot, missing the evolution of perceptions and usage patterns over time. Third, the TAM framework excluded potentially important factors such as trust, social pressure, personal innovativeness, and AI literacy that may significantly influence adoption. Fourth, the lack of qualitative data limited deeper insight into the reasons behind quantitative findings, particularly the unexpected inverse relationships. Fifth, most studies were not blinded, meaning students knew whether they were receiving AI or human feedback, potentially biasing perception measures. Finally, the study focused on quantifiable outcomes that may have overlooked important qualitative dimensions of AI effectiveness, such as emotional support, relationship-building, and the full breadth of value that human interaction provides in educational contexts.
Future Directions: The researchers recommend several directions for future research. First, employing mixed-methods approaches combining quantitative surveys with qualitative interviews or focus groups would provide deeper insight into why perceived ease of use may negatively relate to satisfaction and continuance intention. Second, longitudinal studies tracking students' perceptions and usage patterns over extended periods would capture evolving trends and long-term impacts. Third, expanding TAM to include additional constructs such as trust, social influence, personal innovativeness, perceived risk, AI literacy, and cultural learning expectations would provide a more comprehensive understanding of adoption dynamics. Fourth, future research should examine cost perceptions, perceived value propositions, and economic barriers more deeply, particularly regarding subscription models and equitable access. Fifth, studies should include more diverse samples across geographical regions (beyond Metro Manila), educational levels (K-12, higher education, professional training), and academic disciplines to enhance generalizability. Sixth, investigating academic integrity concerns, authenticity anxieties, and over-reliance issues through empirical research would address critical ethical dimensions. Finally, comparative studies examining differences between developing and developed educational contexts would illuminate how technological infrastructure, digital literacy, and cultural factors shape AI adoption patterns differently across settings.
Title and Authors: "Acceptance of and Satisfaction with ChatGPT as Learning Tool in Higher Education Through the Technology Acceptance Model (TAM): A Case from the Philippines" by Louie Giray, Bench Fabros, Joseph Villarama, Daxjhed Louis Gumalin, and Jeanyxia Bea Saavedra.
Published On: August 8, 2025
Published By: Internet Reference Services Quarterly (Routledge, Taylor & Francis Group)