One of the growing technologies around the world today is Artificial Intelligence (AI). AI already emerged as a dominant and transformational strength. The foray of AI into education can be attributed to the development of early chatbots in the 1960s. (Weizenbaum, 1966)in his research developed the ELIZA programme and tested it on secondary students (Strzelecki, 2023) . The result was that the programme could solve the given problems. Over time, great improvements in AI capabilities, in the area of generative AI, have become apparent in the last decade or so. The rapid evolution has led to AI evolving into a new content- generating thing or what can be dubbed generative artificial intelligence (GenAI) that can change the education landscape

GenAI can generate various types of content similar to human output like text, images, video, and audio that is able to answer instructions given in raw text (Wang et al., 2024). By proposing innovative tools, there is potential for the education space to enhance teachers‘ pedagogical practices and students’ learning performance in the classroom. Today, GenAI has driven great advancements in the field of education and brought about major changes in the way learning is done. GenAI tools, for example ChatGPT, can give valuable assistance to students by giving useful support and improving their adaptable and highly interactive learning environment. (Baidoo-Anu & Ansah, 2023)(Baskara et al., 2023)

But based (Celik, 2023), still many educators are yet to realize its potential in teaching and learning. Although there are calls for greater application of AI, several educationists resist using software that gathers student data or remain skeptical of IT companies heralding the technology as the answer to all issues in education areas (Stockman & Nottingham, 2022). The introduction of GenAI technologies such as ChatGPT in education has the power to revitalize old-fashioned education techniques, enhancing the learning experience and promoting soft skills development for its adopters.

ChatGPT or Chat Generative Pre-Trained Transformer as an AI tool released by OpenAI, on 20 November 2022, a webassisted chatbot and is free to users who have an OpenAI registered account. A chatbot is AI-based software which is designed to simulate human speech over audio or text, and delivers personalized natural language responses to human input in a spoken setup. Even while the predictive linguistic tech work behind ChatGPT is readily applicable since at a less in 2018, ChatGPT provides an easily accessible and easy to utilize platform using both established and newer AI technologies (Boscardin et al., 2024). According to a survey conducted by Populix, in (Annur, 2023), ChatGPT is the most widely used AI application in Indonesia. Noted, as many as 52% of respondents have used the generative AI platform. No wonder natural language capabilities, multilingual support and multifunctionality that are both understandable and increase the productivity of various groups including students make ChatGPT the choice of most people. This is where the researcher made ChatGPT as a generative AI that was researched in the study.

Many university students first discovered ChatGPT via social media platforms (Bonsu & Baffour-Koduah, 2023). The utilization of ChatGPT to enhance learning, for example ChatGPT may be utilised as a tool to create an answer to a theory-orientated query and produce a starting thought for an essay ((Kasneci et al., 2023); (Alafnan et al., 2023)), However, students also have to realise the need to check the validity of the responses produced by ChatGPT. ChatGPT's advanced conversational capabilities can provide constructive feedback on essays and serve as a mentoring mechanism by encouraging crucial thought or discussion between students. Research by (Ansari et al., 2024) shows that the use of ChatGPT is an easy tool to use by students, teachers, and researchers to assist their tasks in their role as user assistants. Students who used ChatGPT were categorized into three main purposes: operationalizing learning, writing assistance, and adaptive learning. The outcomes of the research show that ChatGPT simplifies and describes complicated concepts in an easy-to-read way, that offers students the ability to access learning opportunities that assist in making the student's understanding clearer. ChatGPT works as a tool to familiarise students with unknown concepts and tasks and acts as a platform for continuous learning in comprehensive course subject domains.

Another study focused on students' utilisation of ChatGPT, which is helpful in the writing of essays or tasks, their work in courses and tutorials, and as an analysation device in many academic departments (Rasul et al., 2023). (Strzelecki, 2024) researched university students' acceptance of ChatGPT, finding that habit and performance expectations most affect behavioural intentions towards ChatGPT usage. (Dahri et al., 2024) found high acceptability of ChatGPT in university students which revealed that student acceptance is influenced by many factors, one of which is social influence.

To understand how ChatGPT is utilized by students learning, the researcher chose a well-tested model of technology acceptance and utilization. Unified Theory of Acceptance and Use of Technology or UTAUT 2 model incorporates insights from basic models of acceptance and utilization of information technology. UTAUT 2 takes into consideration several key variables that affect an individual's behavioral intention and their actual usage of a specific technology system (Venkatesh et al., 2012). These factors include Habit (HT), Performance Expectancy (PE), Facilitating Conditions (FC), Effort Expectancy (EE), Price Value (PV), Social Influence (SI), Hedonic Motivation (HM), Behavioral Intention (BI), and Use Behavior (UB). In addition, there are three moderating variables in the UTAUT 2 model, namely gender, age, and experience. The UTAUT 2 model has been used in research exploring a range of educational technologies e.g. digital tools (Hoi, 2020), learning systems (Zacharis & Nikolopoulou, 2022), and learning management software (Raza et al., 2022).

Implementation of UTAUT 2 in this research is expected to facilitate the understanding of student involvement and acceptance models on ChatGPT, resulting in increased knowledge in this newly existing research area. However, the appearance of ChatGPT is causing concern about its potential impact, especially at the level of university. As research by (Abbas et al., 2024) showed that students who were experiencing high academic related workload and time stress to accomplish their task reported greater use of ChatGPT. In addition, students who frequently used ChatGPT were more procrastinating than those who used the tool frequently. Likewise, research (Lo, 2023). The effects of ChatGPT on university education exposed challenges relating to the possibility of informational errors and student plagiarisation.

This methodological approach aligns with the approach recently utilised in a study investigating factors influencing the acceptance and use of ChatGPT among university students. The research reported in that study corroborated the appropriateness of (Strzelecki, 2023)) adapted UTAUT2 model for understanding the use of ChatGPT among the university student community identify Habit as the most important construct affecting behavioural intentions, along with Hedonic Motivation and Performance Expectancy.

Research on the use and acceptance of ChatGPT in education is still small because the launch of ChatGPT is also arguably the latest worldwide (Habibi et al., 2023), including in developing countries such as Indonesia. Even though ChatGPT itself has entered into Indonesian education in meeting the learning needs of student (Maulana et al., 2023). Therefore this research is important to provide an understanding of the application of ChatGPT in education, especially in Indonesia. The application of the UTAUT 2 model in the field of higher education is also still little researched in Indonesia. This study is intended to contribute insights and knowledge about the factors that influence behavioural intentions in using technology, specifically ChatGPT in student learning. This research can contribute to research on the use of UTAUT 2 in the realm of learning and AI technology. As well as being able to provide information for policymakers such as universities and educational institutions in implementing technology in learning. For students, this research can support the adoption of ChatGPT in learning which can maximize learning in the classroom as well as improve students' digital skills. The novelty of this This study centres on the usage of ChatGPT in student learning which is measured using the UTAUT 2 model which involves all variables that affect the use of ChatGPT without any exceptions. In contrast to ((Zacharis & Nikolopoulou, 2022); (Purbonuswanto et al., 2024)); (Gansser & Reich, 2021) research which only examined some variables from UTAUT 2 or without involving moderators of age, gender, and experience. This study aims to find out how the use of ChatGPT in student learning by using variables in the UTAUT 2 model. To analysing the extent to which the variables in UTAUT 2 affect the usage of ChatGPT on college students' learning. Although many international studies have examined the acceptance of ChatGPT in student learning using UTAUT2, in Indonesia itself the research is still relatively new. Moreover, the use of UTAUT 2 in previous studies only focused on certain variables and even without involving moderators. So, from this gap, this research needs to be carried out in Indonesian education.

The operational definition of each varia- ble is used to better understand the meaning of each variable that produces the hypothesis in this study shown in the Table 1.

This research uses a quantitative survei design where data is obtained from a questionnaire using a Likert scale of 1-5, with a cross sectional approach, where data is collected at one time by respondents who are representative of a particular population. Researchers selected this method as it is efficient in collecting large-scale data and allows statistical analysis of variable relationships. The UTAUT 2 model is relevant because it is able to explain the variables that influence intention and behavior in the use of technology. The research model is illustrated in Figure 1.

The sampling technique is simple random sampling, where everyone in the population has the same opportunity to be selected as a respondent. this technique can avoid bias and ensure that the selected sample can represent the population. as for the population selection criteria in this study are students of the Surabaya State University Economics Education Study Program. This is based on the level of technology use in students who are arguably frequent, especially on ChatGPT in their learning. The population amounted to more than 500 students and the sample calculation used the formula (Hair et al., 2016).

N = (5 - 10 x number of indicators) N = 10 x 34

N = 340

That research took a sample of 340 students from the Surabaya State University Economics Education Study Program. The instrument used is made based on the indicators in each variable in the UTAUT 2 displayed in the Table 2 each item was adapted to reflect the context of ChatGPT usage in learning. To ensure the validity and reliability of the research instruments, validi- ty and reliability checks were performed uti- lizing the SmartPLS PLS-SEM approach. The validity test is conducted by examining the outer loading and AVE, while the relia- bility test examines Cronbach alpha and composite reliability.

The data obtained will be calculated using the Partial Least Squares Structural Equation Modeling (PLS-SEM) technique. The analytics will be processed with SmartPLS 4.0 software. PLS-SEM is a causal modeling based approach that maximises the variety of latent variables and is capable of working efficiently using small samples and complex models. SEM-PLS analysis can analyse unobserved variables (variables that cannot be directly measurable), test complex research models, and take into account measurement error in a test (Memon et al., 2021) which has several tests, namely the outer model and inner model (Farida et al., 2022). Testing the outer model provides a value for validity and reliability analysis. This technique is very suitable for complex theoretical research such as UTAUT 2. The advantage of PLS - SEM is that it is able to analyze data even though the sample size is small, the data distribution is not normal, and it is still able to provide strong analytical value.

Data collection was conducted from the results of distributing Google forms online was carried out in March 2025. The questionnaires that have been filled in by Economics Education students at Surabaya State University totaled 340 respondents. The results of the questionnaire collected Table 3.

The information in the table shows the classification of respondent characteristics including gender, age, length of use, and frequency of use of ChatGPT. It was summarised that the majority of students who became respondents were female student. While the majority of ChatGPT user respondents' age is 19. The majority of respondents have used ChatGPT for at least 9-16 months and most respondents use it several times a week.

After analyzing the characteristics of the respondents, the researcher tested the outer model which gave a value to the validity as well as reliability analysis. In the outer mod- el, there are several requirements, namely outer loading > 0.7 and Average Variance Extracted (AVE) > 0.5 in validity analysis, while Cronbach Alpha and Composite Relia- bility> 0.7 in reliability analysis (Hair et al., 2020). The outer model results in the SEM-PLS test on SmartPLS

4.0 are depicted in the model at Figure 2.

Each instrument has a value of > 0.7 which means it has met convergent validity. While the value of AVE, Composite

Reliability, and Cronbach's Alpha could be viewed in Table 4.

The AVE value in every variable is > 0.5 which shows that convergent validity is met and can be said as valid. All variables have a Cronbach's Alpha and Composite Reliability value > 0.7 which indicates that all variables have been fulfilled and can be considered reliable.

In the inner model, can define the relationship among existing latent variables. This model describes the relationship structure based on existing theory. The testing of the structural model is done in several ways, namely on the R Square value to assess the goodness of a model. There are three more criteria for the strength of R Square, which are 0.67 is strong, 0.33 is medium, and 0.19 is weak. The resulting R Square test is shown in theTable 5.

The information in Table 4 present that the R square on ChatGPT user behavior is 0.553, which means that Behavior Intention has an effect of 55.3%, meaning that the percentage of behavior intention variables can be described by Social Influence (SI), Performance Expectancy (PE), Facilitating Conditions (FC), Effort Expectancy (EE), Hedonic Motivation (HM), Habit (H), and Price Value (P) while the remainder 44.7% is are described by another variable that is not included in this research. The variable use behavior may be described by Behavior Intention of 0.517 or 51.7% while the remaining 48.3% is described by other variables not included in this research. Both R Square values are classified in the moderate or medium category.

Furthermore, hypothesis testing is carried out which is seen through the original sample value, p-value, and F Square (Khan et al., 2022). The moderating effect test is seen from the F Square value which is divided into three categories, namely 0.005 is categorized as low, 0.01 is categorized as medium, 0.025 is categorized as high, while the direct effect has a different category range, namely 0.02 is categorized as low, 0.15 is categorized as medium, and 0.35 is categorized as high. (Hair et al., 2019). The results of the test are displayed in Table 6

From the results of hypothesis testing, it shows that the variables that have a significant affect behavioral intention to use ChatGPT are performance expectancy (PE) moderated by age, effort expectancy (EE) moderated by gender, social influence, and habits. Then the three variables, namely facilitating conditions (FC), habit (HT), and behavioral intentions (BI) are proven to affect use behavior (UB). The moderating variable of age only significantly strengthens performance expectancy (PE) and the moderating variable of gender significantly strengthens effort expectancy (EE) in affecting user behavior intention (BI) of ChatGPT. While experience is not proven to moderate any variable at all.

Hypothesis testing results show that performance expectations (PE) moderated with age affect the behavioral intention (BI) of using ChatGPT on students in learning but not on gender and experience. This illustrates that age can strengthen belief in the benefits of ChatGPT in student learning regardless of gender or experience of using ChatGPT. This result is supported by research which states that PE has been proven to have an effect on BI because ChatGPT does promise productivity for individuals ((Bazelais et al., 2024); (Rahmawati et al., 2022)). However, it is contrary to (Strzelecki, 2024) research which states that performance expectations (PE) moderated by age have no effect on behavioral intentions (BI). Although both are in the field of education, the age generation of the selected population can also make research results different. This shows the need for further research to explore these variables more deeply.

Similarly, Effort Expectancy (EE) is proven to influence Behavioral Intentions (BI) of ChatGPT use by students in learning moderated with gender but not age and student experience. This illustrates that differences in student gender can strengthen assumptions about the ease of using ChatGPT in their learning. The effect is in line with the research (Strzelecki, 2023)) but contradicts research (Alshammari & Alshammari, 2024) which found that gender moderated the relationship. Given these differences in findings, further research is needed in the future on the factors that influence the use of ChatGPT in student learning.

Habit (HT) and Social Influence (SI) also influence the Behavioral Intention (BI) of using ChatGPT on student learning regardless of gender, age, and experience of using ChatGPT on students. These results are in line with the research (Tao et al., 2024) and (Osei et al., 2022) but contradict research (Zacharis & Nikolopoulou, 2022). This difference could be due to the characteristics of the respondents studied given the different places of the study. Therefore, the researcher hopes that this research will be able to contribute and at the same time invite further researchers, especially in Indonesia, to contribute to studying the adoption of ChatGPT in student learning.

While the three factors, namely facilitating conditions (FC), habit (HT), and behavioral intentions (BI) are proven to influence use behavior (UB), but the three moderating variables are not proven to strengthen this influence. The Declined hypothesis shows that there is no influence on the variables of facilitating conditions (FC), price value (PV), and hedonic motivation (HM) on behavioral intention (BI) using ChatGPT in student learning ((Gansser & Reich, 2021); (Grassini et al., 2024))

The moderating variables of experience, gender, and age in this research mostly did not moderate the tested relationship. This result is in keeping with the results of (Strzelecki, 2023) The absence of a significant effect on this moderating variable may be due to the ease of use on ChatGPT which can be accessed regardless of age, gender, and experience in students who should be the latest in technology including in the field of learning.

This research has an implication for educators and stakeholders in developing strategies for adopting technology in teaching. For example, technology training, facility support, and supportive policies that suit the needs of today's students. This insight can create an adaptive and innovative learning environment. Several rejected hypotheses are the shortcomings of this study. So it is hoped that further research will further examine the acceptance of ChatGPT in student learning with a wider and more diverse sample and a more general context. As well as using a more appropriate methodological approach in order to gain a broader and deeper understanding of the study.

This study analyzes the usage of ChatGPT in student learning using factors in UTAUT 2 which are considered capable of measuring adoption and utilisation of technology in individuals. The findings showed that performance expectancy moderated by age, effort expectancy (EE) moderated by gender, social influence (SI) and habit (HT) influenced behavioural intention (BI) of using ChatGPT in student learning. In the variable facilitating conditions (FC), habits (HT), and behavioral intentions (BI) are proven to affect use behavior (UB) but the three moderating variables are not proven to strengthen this influence. While the variables of facilitating conditions (FC), price value (PV), and hedonic motivation (HM) show no influence on behavioral intention (BI) using ChatGPT on student learning. The moderating variables in this study mostly did not moderate the tested relationship. Hopefully, this study makes a theoretical contribution by testing the relevance of the UTAUT2 model in the context of ChatGPT adoption and identifying the key factors that affect the intention and behaviour of users. The present findings indicate that learning environment and habits influence the utilisation of ChatGPT in student learning. Therefore, it is recommended that educational institutions provide training and technical support to encourage optimal utilisation of ChatGPT. Given that this research is limited to UTAUT2 variables, future studies are expected to explore additional factors such as digital literacy and cultural context to broaden the understanding of technology adoption in education.