Menon & Shilpa [4] explain that ChatGPT is an advanced language model developed by OpenAI, first released in 2018 and based on the Generative Pre-trained Transformer (GPT) architecture. It can produce grammatically accurate, human-like responses to text inputs because it was trained on a large amount of textual data. ChatGPT is designed for conversational applications and generates highly realistic responses that demonstrate a deep understanding of context. Its applications include virtual assistants, customer service, language translation, and personalized tutoring. However, significant concerns regarding data privacy, decision-making transparency, and data quality must be addressed before widespread adoption in academic and research environments can occur.

A qualitative study by Raman et al. [5] states that OpenAI launched ChatGPT as its state-of-the-art AI language model in November 2022. The platform reached one million users within its first five days, followed by 100 million users in three months, setting a new record for technology adoption. The model gained widespread recognition through publications in scientific journals and mainstream media. The rapid adoption rate of ChatGPT highlights its importance in natural language processing, as it has surpassed the growth trajectories of major social media platforms. The GPT-3-based system demonstrates superior human-like response generation capabilities despite the educational and creative concerns associated with its use in various business applications.

The authors Alshammari & Alshammari [6] describe ChatGPT as an AI chatbot developed by OpenAI in November 2022, which gained popularity due to its human-like text generation abilities achieved through processing large datasets using machine learning algorithms. The potential of ChatGPT to transform education, teaching, and assessment methods is undeniable; however, its effects on colleges, particularly concerning assessment practices and students’ writing and critical thinking abilities, remain a point of concern. Although the academic applications of ChatGPT are promising, researchers have not yet adequately examined how students adopt and utilize this technology.

Recent research in the Gulf region has begun to explore the relationship between ChatGPT and academic performance. Al-Mamary et al. [7] applied the technology-to-performance chain theory to investigate Saudi Arabian university students’ use of ChatGPT and found that task characteristics, technology characteristics, and individual characteristics influence task-technology fit, which subsequently affects ChatGPT utilization and academic performance.

The implementation of ChatGPT within educational settings offers multiple advantages that benefit university students through improved academic outcomes. Tajik & Tajik [8] and Rasul et al. [9] both highlight that the adoption of ChatGPT has substantial potential to enhance student performance. The platform provides two major benefits: it reduces barriers for students who speak English as a second language and supports students with disabilities. The system includes inclusive features that generate effective responses for visually impaired learners, create concise summaries for students with learning difficulties, and translate instructional materials to align with students’ language abilities.

Tajik & Tajik [8] also describe how ChatGPT assists students by offering relevant and engaging learning sessions and responding to their questions. This tool can be used to organize and manage scholarly literature and to aid in memorization of information learned during conversations. Firat [10] also agrees that the supplementary materials produced by ChatGPT are very helpful for academic work and research. It can help to improve the writing skills of the students by producing activity summaries.

Mhlanga [11] stated that ChatGPT is a flexible tool that allows students to get learning materials at any time and from any place. This will help students to become independent in their studies and, at the same time, teach them how to maintain a balance between studies and other activities. ChatGPT provides a personalized learning experience that helps students in understanding the course material and also assists them in solving some learning challenges. The personalized approaches in ChatGPT also enable the students to connect and thus create an integrated learning environment.

Many people have benefited from using ChatGPT, but there are many downsides and negative effects to relying heavily on technology and possibly diminishing students’ ability to think critically.

Goodhue and Thompson [12] proposed the Task-Technology Fit (TTF) Model in 1995 to explain the level of technology adoption on the basis of the fit between a technology and the tasks and needs of the users. Marikyan and Papagiannidis [13] stated that the TTF Model identified that there is a knowledge gap related to the impact of technology usage on individual performance. The three models of TRA, TPB, and TAM have limitations as they concentrate on the perceived usefulness and attitude of the technology and not the individual performance and actual decision-making. The TTF Model suggests that better performance will result from a good task- technology fit, a claim supported by research on the level of task-technology fit and technology use.

The Technology-to-Performance Chain (TPC) Theory, also developed by Goodhue & Thompson [12] in 1995, expands the rationale for technology implementation by explaining how technology supports task performance. TPC focuses on both the nature of the tasks and the suitability of the technology used. This theory is particularly useful for understanding how ChatGPT can support college students in various learning activities through interactive exercises and instructional materials.

The TAM model incorporates task-technology fit as a performance indicator and integrates user attitude as a predictor of use. Nisak et al. [14] and Mbangata & Abayomi [15] state that task-technology alignment is influenced by task features, technology attributes, and individual characteristics. This alignment directly affects performance and indirectly affects technology use through factors such as expected outcomes, social norms, and facilitating conditions.

The initial study by Goodhue and Thompson demonstrated how task-technology fit influences performance outcomes, and the model has since been applied in fields such as software development, administrative decision-making, and healthcare. However, the application of TPC Theory in e-learning and ChatGPT implementation remains an underdeveloped research area. The TPC framework provides academic researchers with an opportunity to examine ChatGPT performance variables, given its essential role in explaining technology-to-performance relationships.

The Technology-to-Performance Chain (TPC) framework is a useful structure for analyzing and influencing intentions to use ChatGPT. The first step involves examining task characteristics, technological characteristics, and individual characteristics to assess readiness for and adoption of ChatGPT. These attributes provide a holistic view of user orientations and needs. Task-Technology Fit ensures that ChatGPT is well-matched to organizational requirements and functions efficiently. The final component, Utilization, emphasizes the value of actual use by highlighting the adoption and effective implementation of ChatGPT across the organization.

The research by Al-Mamary et al. [7] supports this approach, as they applied the technology-to-performance chain theory to study ChatGPT adoption in educational settings and found that task characteristics, technology characteristics, and individual characteristics determine task-technology fit and subsequent performance outcomes.

The Technology Acceptance Model (TAM) is one of the most widely used models for predicting individual users’ acceptance and use of information systems and technology. Surendran [16] identifies TAM as a leading model that explains technology adoption, while Agrawal [17] emphasizes that it is one of the most frequently applied frameworks in studies on IS/IT acceptance. The individual-level IS adoption prediction model known as TAM was developed by Davis [18]. TAM consists of two essential components: perceived ease of use and perceived usefulness.

Davis [18] defined perceived usefulness as the degree to which a person believes that using a system will improve their job performance. A system is more likely to be used if the individual believes it will help them perform their work more effectively. This variable captures the user’s belief in the system’s ability to enhance their performance. Perceived ease of use, on the other hand, refers to the extent to which the user believes the system is easy to operate or how much effort they think is required to use it [18].

The TAM model illustrates how attitude influences user intentions according to its design. Users form their attitudes toward a system based on perceived ease of use and perceived usefulness, which directly shape their behavioral intentions. When users believe a system is both useful and easy to use, they develop positive attitudes that increase their intention to adopt it. Individuals with positive attitudes toward a system demonstrate stronger intentions to use it, which ultimately leads to actual system usage.

This research introduces a comprehensive conceptual framework to analyze the factors influencing individuals’ willingness to use OpenAI’s advanced AI-powered chatbot, ChatGPT. Five main independent variables form the basis of this model: Task Characteristics, Technology Characteristics, Individual Characteristics, Task-Technology Fit, and Attitude.

Users evaluate ChatGPT through two key dimensions. Task Characteristics represent the nature and difficulty level of the tasks they intend to perform, while Technology Characteristics reflect their perceptions of the chatbot’s qualities and capabilities. Individual Characteristics encompass users’ backgrounds, personal perspectives, and traits that shape their willingness to adopt the technology. Task-Technology Fit assesses how well users’ tasks align with the features and functions offered by ChatGPT. Attitude toward using ChatGPT indicates users’ overall positive or negative evaluation of the tool based on its perceived usefulness and ease of use.

This model examines these variables in detail to provide valuable insights into ChatGPT adoption intentions, offering benefits for researchers, practitioners, and policymakers working on AI adoption and implementation. Figure 1 presents the study model.

To investigate university students’ intention to adopt ChatGPT, five hypotheses were developed based on the Technology-to-Performance Chain (TPC) and Technology Acceptance Model (TAM). Each hypothesis is supported by relevant literature, as explained below:

Task characteristics refer to the nature, complexity, and requirements of academic tasks students engage in. When students perceive that a technology tool helps them accomplish their academic tasks efficiently, their likelihood of adopting it increases. Fang et al. [19] and Alcover and Topa [20] found that task relevance and clarity significantly influence individuals’ acceptance of technological tools.

H1: Task characteristics have a positive and significant effect on students’ intention to adopt ChatGPT.

Technology characteristics include usability, reliability, and compatibility with students’ learning tools. Students who perceive ChatGPT as easy to use and reliable are more inclined to adopt it. Aziz and Wahid [21] and Mohamad et al. [22] confirmed that positive technology perceptions are strong predictors of adoption behavior.

H2: Technology characteristics have a positive and significant effect on students’ intention to adopt ChatGPT.

Individual characteristics reflect students' personal traits, including digital literacy, openness to new technologies, and self-efficacy. Students who are comfortable with new tools and possess higher digital competence are more likely to use AI-based systems. This is supported by Chadegani et al. [23] and Tabak and Barr [24].

H3: Individual characteristics have a positive and significant effect on students’ intention to adopt ChatGPT.

Task-Technology Fit (TTF) refers to the extent to which the features provided by ChatGPT are aligned with the tasks of students. A higher TTF leads to higher perceived usability and intention to use the technology. Alyoussef [25], El Said [26], and Lu and Yang [27] all agreed that technologies are more likely to be adopted when there is a higher fit between the tasks and the technology.

H4: Task-Technology Fit has a positive and significant effect on students’ intention to adopt ChatGPT.

Attitude reflects students’ overall perception and emotional response to using ChatGPT. Positive attitudes increase the likelihood of behavioral intention. This relationship is strongly supported by Abdaljaleel et al. [28], Acosta-Enriquez et al. [29], and Haq et al. [30].

H5: Attitude toward using ChatGPT has a positive and significant effect on students’ intention to adopt the tool.

The main goal of this survey research is to determine what factors affect college students' adoption of ChatGPT. The study collected data on task characteristics, technology characteristics, individual characteristics, task-technology fit, attitude, and intention to adopt ChatGPT using validated survey instruments. The proposed study model was validated using Structural Equation Modeling (SEM) with AMOS and SPSS.

This study employed a non-probability convenience sampling technique, which is commonly used in behavioral and educational research where access to a complete sampling frame is not feasible. The researchers distributed an online questionnaire via social media platforms and university student groups. This method allowed efficient access to a large and diverse sample of students within a limited timeframe and budget.

Participants were included in this cross-sectional study if they were currently enrolled students at Al-Razi University (Yemen) during the data-collection period, aged 18 years or older, and had prior experience using ChatGPT or similar AI text-generation tools, enabling them to evaluate the study constructs. Only students who understood the survey instructions, provided informed consent, and completed all mandatory questionnaire items were retained in the final dataset.

Participants were excluded if they were younger than 18 years, not enrolled at Al-Razi University, had no prior exposure to ChatGPT, declined participation, or did not provide consent, or submitted incomplete/invalid questionnaires or duplicate responses, which could not be used for SEM analysis.

To mitigate sample bias, the survey link was disseminated across multiple academic departments, student clubs, and administrative channels to encourage participation from students across different majors, levels of study, and backgrounds. The research team emphasized inclusivity by encouraging both male and female students to participate and by ensuring the invitation reached a wide age range (18-50 years). The anonymity of responses was maintained to reduce social desirability bias, and participation was voluntary to ensure authentic engagement.

The sample size of 209 respondents is justified based on the requirements of Structural Equation Modeling (SEM). According to Kline [31], a sample size of 200 or more is generally adequate for SEM when the model includes multiple constructs and paths. Additionally, it has been recommended that a minimum sample size of 5-10 observations per estimated parameter is needed; the current sample size meets this requirement given the number of constructs and observed indicators used in the model.

A post hoc power analysis was conducted using a medium effect size (0.30), a significance level of 0.05, and a sample size of 209. The analysis showed an actual power of 0.87, confirming that the sample size was sufficient to detect statistically significant relationships in the model.

Therefore, the sample provided adequate statistical power for model estimation and hypothesis testing, while also offering insights into university student perspectives on AI tool adoption within the Yemeni higher education context.

All items in the questionnaire were adapted from previously validated instruments to suit the context of ChatGPT adoption in education (Appendix A). The items were reviewed by academic experts for content validity and adjusted linguistically for clarity and relevance. Each construct was measured using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree). Table 1 presents the item statements, adapted sources, and standardized factor loadings.

The authors collected data using online surveys from a sample of 209 participants. Regardless of their academic program, level of study, or location, all students enrolled in a private university in Yemen (institutional name anonymized) were encouraged to participate in the questionnaire. Participants were required to be between 18 and 50 years of age. The demographic profiles are presented in Table 2.

The demographics of the respondents are presented in Table 2. The results show that 56.94% of participants were men, while 43.06% were women. Furthermore, 93.78% of participants were under the age of 30, 3.83% were between 31 and 40 years old, and 2.39% were between 41 and 50 years old. Regarding educational level, 18.66% of participants held a diploma degree, 78.95% held an undergraduate degree, and 2.39% held a postgraduate degree.

In terms of ChatGPT familiarity, most students demonstrated a high level of ability, with 67.46% reporting that they could use ChatGPT for learning purposes. Additionally, 29.19% of respondents indicated a moderate ability to use ChatGPT for learning objectives, while only 3.35% reported difficulty using ChatGPT for educational purposes.

The confirmatory factor analysis (CFA) from AMOS version 24 was used to test the measurement models for each construct [32]. Awang [33] suggested that CFA should be used to assess whether the measurement items reflect the researcher's conceptualization of the construct. The measurement models were determined to satisfy the requirements suggested by Al-Mamary & Alshallaqi [34], as shown in Fig. (2).

The study employed average variance extracted (AVE) to assess scale validity. The AVE value needed to meet or exceed 0.5, as recommended by Shrestha [35], Abdulrab et al. [36], Al-Mamary et al. [37], and Mutahar et al. [38], to ensure convergent validity (Tables 3 and 4). The Fornell-Larcker criterion was used to examine the discriminant validity of the measurement model. The square root of the Average Variance Extracted (AVE) for each construct should be greater than the correlation between that construct and all the other constructs.

As a research tool, Structural Equation Modeling (SEM) can enable researchers to model multiple variables/ components. A construct can only be included in an SEM study after rigorous testing to confirm its unidimensionality, validity, and reliability [34].

The evaluation of model fit and relationship strength relies on goodness-of-fit indices. When a model demonstrates a good fit, the data support the relationships among variables. The essential component of SEM, presented in Figure 2, enables researchers to test specific hypotheses regarding relationships between model variables, thereby helping to validate or reject the research questions and hypotheses.

The development and interpretation of SEM require extensive knowledge of the research domain as well as advanced statistical expertise. The method provides reliable results for examining complex relationships between multiple variables, producing useful insights for both theoretical advancement and practical application.

The proposed conceptual model in Figure 3 demonstrates strong goodness-of-fit indices, indicating its ability to explain the relationships between Task Characteristics, Technology Characteristics, Individual Characteristics, Task Technology Fit, attitude, and intention to adopt ChatGPT. The complete results from the hypothesis testing are presented in Table 5.

The research results will be useful for educational staff members, government officials, and technology system developers who aim to enhance the use of AI in colleges. Identifying the factors that influence students' willingness to use ChatGPT will help universities develop appropriate strategies for integrating AI-based tools. Educational institutions should establish AI-friendly policies that encourage responsible use while addressing ethical concerns related to academic integrity and plagiarism. Educators should utilize ChatGPT to create learning experiences tailored to individual student needs, offering a more engaging and dynamic learning environment.

The research also highlights the necessity of developing digital infrastructure and training programs for both students and faculty members. Universities should implement workshops and training sessions that help students acquire the skills needed to maximize the benefits of AI-assisted learning. Developers should enhance AI tools by adding features that better serve students, making them more accessible, interactive, and supportive of different languages.

The deployment of AI educational tools requires policymakers to establish regulations that ensure transparency and reliability. As AI technologies expand, policymakers should create appropriate standards that protect academic integrity while fostering innovation. This research provides specific recommendations for developing an organized framework for AI implementation in education, which will enrich and support traditional learning approaches.

The research contributes to the technology adoption literature by applying the Technology-to-Performance Chain (TPC) theory and the Technology Acceptance Model (TAM) within AI education settings. The study confirms the importance of task characteristics, technology characteristics, and individual characteristics in determining technology acceptance by examining their effects on students' adoption intentions.

The study further contributes to the literature by incorporating Task-Technology Fit (TTF) as an independent variable, offering a more detailed understanding of how AI technologies align with students' learning preferences. It also reinforces the role of attitudes toward AI, highlighting the significance of user perceptions in shaping behavioral intentions.

The findings demonstrate the expanding role of AI technology in education and complement previous studies on ChatGPT implementation in academic environments. This research enhances existing knowledge by integrating ChatGPT into an established theoretical framework, thus strengthening the predictive capabilities of technology adoption models. Future studies should build upon these results by examining how cultural differences, digital literacy, and institutional support influence the adoption process.

Overall, this research extends theoretical knowledge on AI adoption and provides foundational elements for future studies on AI-based educational technologies, offering a systematic framework to analyze their effects on learning outcomes.