Enhancing Learning Beyond Correction: AI-Assisted Japanese Business Writing and Sociocultural Awareness in Online Higher Education

Abstract

Artificial intelligence (AI) is rapidly transforming language education. However, its pedagogical and sociocultural impacts on Japanese business writing remain underexplored. This study aims to examine how ChatGPT4o-based automated feedback functions within Japanese business writing education for adult learners in online higher education, with particular attention to both its instructional impact and learners’ sociocultural awareness. Situated in a cyber university context where the proportion of adult learners is increasing, the study explores the potential of AI-mediated feedback to address learners’ diverse educational and cultural needs. It employed a mixed-methods design, combining a survey of 27 participants and in-depth interviews with 11 participants. The interviews were transcribed and thematically coded to gain deeper insights into learners’ perceptions. The findings indicate that ChatGPT feedback contributed to learners’ planning of study strategies, the provision of immediate and personalized corrections, the reinforcement of error awareness, and the acquisition of honorific and polite expressions. On the one hand, learners reported that they could quickly understand regional language practices and communication conventions in business contexts, thereby deepening their cultural sensitivity. On the other hand, some learners expressed concern that increased reliance on AI could weaken exploratory and critical learning. These results suggest that ChatGPT can serve not merely as a correction tool but also as an educational resource that simultaneously fosters self-directed learning and sociocultural competence. However, to ensure reliability and cultural appropriateness, hybrid feedback incorporating instructor guidance is necessary. This study has academic significance in demonstrating the potential of extending AI-based feedback to Japanese business communication education, thereby constructing an integrated language and culture learning environment.

1. Introduction

Many scholars have long noted the difficulty of Japanese business writing (Jenkins & Hinds, 1987; Kameda, 2014; Yuniarsih & Putra, 2023). because it is characterized by the inclusion of cultural elements, such as seasonal greetings, and by the prominent use of honorific expressions. Due to these complex factors, learning Japanese business writing requires sufficient practice based on various situations and formats (Awamura, 2010). However, as Riddell (2015) highlights, to improve their writing ability, it is effective for learners to receive frequent, high-quality feedback, yet it is difficult for instructors to provide feedback continuously (Riddell, 2015). Consequently, attention has increasingly been turned to automated writing feedback (AWF) through artificial intelligence (AI) as a potential solution.

Feedback from AI tools, such as ChatGPT, has been positively acknowledged for providing efficient and comprehensive feedback. This is because it can mitigate the constraints of instructors’ time limitations, lack of resources, and challenges in offering personalized feedback to learners (Wilson et al., 2025). Previous studies have shown that automated feedback not only enhances learners’ linguistic experience but also helps sustain their motivation (Mohammed & Khalid, 2025). However, in domains such as Japanese business writing, cultural context must be included in the feedback, making it difficult to conclude that automated feedback is as effective as it is in more general writing instruction. Moreover, AI-based feedback does not always guarantee positive effects on learners’ outcomes and cannot easily replace the deep insights and affective feedback that instructors provide. Therefore, it is necessary to examine through actual classroom practice whether AI feedback has meaningful educational effects on Japanese business writing.

Against this background, this study analyzed the satisfaction and effectiveness of both AI-based and instructor-provided feedback on the Japanese business writing of cyber university learners who can study without the constraints of time and place. Learners at cyber universities tend to show high levels of self-directedness and prefer individualized learning pathways (Merriam & Bierema, 2013). Given these characteristics, they are likely to benefit more from both immediate feedback and feedback that can be repeated in response to individualized requests, thereby enhancing their learning efficiency.

Accordingly, this study aims to explore the impact of AI-based automated feedback on cyber university learners of Japanese business writing through an actual instructional case, thereby providing foundational data for developing effective educational strategies using AI tools in online higher education.

The three main research questions (RQs) are as follows:

(1)

RQ1: How does the use of ChatGPT in classroom instruction influence learners’ overall satisfaction and learning experience?

(2)

RQ2: How does ChatGPT-based feedback affect Japanese business document-writing skills?

(3)

RQ3: How does feedback using ChatGPT affect learners’ understanding of Japanese business culture?

2. Literature Review

2.1. Automated Feedback in Language Education

In recent years, automated feedback from AI has drawn increasing attention in the field of language education. Feedback not only informs learners of the goals they must achieve and the tasks they must perform, thereby enhancing learning effectiveness, but also stimulates motivation, influencing goal attainment in various ways (Ilgen et al., 1979). For this reason, systems such as AWF (Fleckenstein et al., 2023; Cotos, 2023), which provides feedback within seconds, and automated essay scoring (AES) (Misgna et al., 2024), which provides scores alongside feedback, have been reported to offer advantages in the field of writing instruction.

Shi and Aryadoust (2024) conducted a representative study that systematically reviewed the status of AWF. They analyzed 83 papers published with Social Sciences Citation Index (SSCI) indexes on AWF from 1993 to 2022 in three aspects—the performance of AWF; the recognition of AWF, use, and participation; and the effect of AWF—all of which showed positive, negative, and mixed results. This indicates that further in-depth research is needed on multiple dimensions, such as accuracy, construct validity, and validity in use, beyond merely adopting AWF. In particular, theoretical standards and empirical evidence remain insufficient regarding whether AWF can completely replace human teachers’ feedback or how it should be integrated as a complementary tool.

Furthermore, Yildiz and Kuru Gönen (2024) investigated the effectiveness of AWE feedback in reducing errors in English writing. Their findings showed that both teacher feedback and AWE feedback were effective in reducing errors; however, AWE feedback minimized errors in mechanics and usage more efficiently, whereas teacher feedback was more often required in the areas of content and organization (Yildiz & Kuru Gönen, 2024). In addition, Yu (2023) reported that AI-generated performance feedback for online learners had a positive impact on their writing practice in online learning spaces, although it was not helpful for creativity (Yu, 2023). Collectively, these studies suggest that, while AWF can have a positive impact on writing outcomes, the type and function of feedback produce different educational effects depending on learner needs.

Meanwhile, studies on learners’ and teachers’ perceptions of AWF provide important implications by clarifying how their actual experiences influence AWF acceptance and future intention to use. Wilson et al. (2024) examined students’ perceptions of usability, usefulness, and preference regarding the MI Write system, an AI-based automated writing assessment and feedback system. Their findings indicated that students with limited English proficiency, from lower socioeconomic backgrounds, and enjoying writing tended to perceive MI Write as relatively more useful (Wilson et al., 2024). Silhadi (2025) reported that Algerian undergraduates majoring in English as a Foreign Language (EFL) used AI tools to generate writing content but were not able to completely utilize them to foster deep learning or develop independent writing skills (Silhadi, 2025). These findings suggest that AWF is received differently depending on individual learner factors, such as language proficiency, interest, and sociocultural context; they point to its potential for complementary use alongside human feedback.

Moreover, technological developments, such as natural language processing, have expanded the range of systems capable of providing AWF. With the emergence of generative AI tools, such as ChatGPT, AWF holds the potential for leveraging large language models (LLMs) to deliver more accurate and comprehensive feedback. Integrating generative AI into AWF creates opportunities to provide deeper and more personalized feedback (Shi & Aryadoust, 2024). In this context, research on AWF that incorporates generative AI is meaningful in that it goes beyond examining the tool’s mere effectiveness to consider AWF’s educational potential as a learner-centered, customized feedback system.

More recently, however, advances in generative artificial intelligence have prompted a reconceptualization of automated writing feedback. Systematic reviews of ChatGPT use in EFL writing contexts indicate a clear shift from rule-based, accuracy-oriented systems toward dialogic and adaptive feedback mediated by large language models (Núñez & Castrillo De Larreta-Azelain, 2023). Empirical studies further suggest that such generative-AI-based feedback is particularly effective when combined with instructor guidance, highlighting the pedagogical value of hybrid feedback models (Asadi et al., 2025). This shift suggests that AI-mediated feedback should be examined not merely as an automated correction mechanism but as a mediational resource that supports learners’ strategic and contextual development in writing education. From a sociocultural perspective, AI-mediated feedback can be understood as a mediational tool that scaffolds learners’ development within specific communicative and cultural contexts.

2.2. Learner Satisfaction in Online Learning

Since the outbreak of COVID-19 and the subsequent spread of online lectures, the importance of learning effectiveness, learner satisfaction, and quality assurance in learning has become even greater. Learner satisfaction is an important indicator for evaluating the quality of the learning experience (Moore & Kearsley, 1996), and as the modes of interaction with instructors and peer learners have changed along with technological advances, research on satisfaction in online environments has become increasingly necessary (Dou et al., 2022).

Various studies have examined the factors that influence learner satisfaction in online learning. Kuo et al. (2013) analyzed predictive relationships centered on interaction, internet self-efficacy, and self-regulated learning as factors affecting learner satisfaction in online learning. Their results showed that learner–instructor interaction, learner–content interaction, and internet self-efficacy were significant predictors of learner satisfaction (Kuo et al., 2013). In addition, Hou and Wong (2024) report that self-efficacy, strong intrinsic and extrinsic motivation, sufficient technological readiness, and a preference for meaningful interaction significantly enhanced satisfaction with e-learning. Interaction positively affects learners’ engagement and satisfaction, and learners tend to perceive courses with active interaction as high-quality classes (Hou & Wong, 2024). These studies underscore the importance of interaction while analyzing satisfaction in online learning from psychological and technological perspectives (Yoon & Kam, 2021).

Meanwhile, some studies have focused on the learner’s lived experience as a dimension of online satisfaction. Jiang (2024), using the case of Chinese university students in Massive Open Online Courses (MOOCs), examined the influence of five dimensions—sensory, emotional, cognitive, behavioral, and social experiences—on user satisfaction in online education platforms. The results showed that all five dimensions had a positive impact on satisfaction, with sensory experience exerting the most pronounced effect (Jiang, 2024). Key factors in sensory experience that determined satisfaction included the platform’s visual design, user interface convenience, and rational information layout.

More recent studies have increasingly attempted to examine the influence of personalized learning pathways. In a study of English as a Second Language (ESL) learners, Zhao (2025) reports that technology-based personalization, including the one provided by AI-based feedback systems, supports self-directed learning and enhances language self-efficacy (Zhao, 2025). Similarly, Zang (2024) constructed a basic framework for student-customized learning modes based on big data technology and, by comparing them with traditional learning modes, found that personalized learning was more effective in terms of knowledge and skills, processes, methods, emotional attitudes, learning achievement, and learner confidence (Zang, 2024). In particular, AI has the potential to create engaging learning environments by providing personalized feedback for performance monitoring and tailoring educational experiences to individual learner needs (Bhatia et al., 2024). Research that takes learners’ personal characteristics into account deserves attention in that it allows for a more nuanced understanding of satisfaction in online learning.

2.3. Business Japanese Education from a Cultural Education Perspective

Sociocultural awareness in business Japanese goes beyond simple cultural knowledge and has been widely discussed as a central component of pragmatic competence in professional communication (Byram, 2021). In business contexts, successful written interaction requires not only grammatical accuracy but also the ability to interpret and respond appropriately to situational factors such as social hierarchy, relational distance, and culturally embedded norms, which are closely related to learners’ pragmatic competence (Ishihara & Cohen, 2014).

In international business, it is necessary to understand unique cultural values, social norms, and business customs to avoid miscommunication (Garcia, 2016). Recently developed curricula and textbooks for business Japanese tend not only to focus on document writing or learning expressions for specific business situations but also to be structured so that learners understand Japanese business culture.

A significant portion of prior research on business Japanese has focused on textbooks. For instance, Hatori (2023) developed business Japanese textbooks applicable to real corporate environments, reflecting various work situations, such as meetings, presentations, and email writing, while providing explanations of cultural backgrounds to deepen learners’ cultural understanding (Hatori, 2023). Similarly, Kaname et al. (2021) state that university-level business Japanese courses should aim to cultivate basic social competence and propose the publication of textbooks integrating business situation simulations, case-based learning, and self-reflection activities (Kaname et al., 2021). This integrated approach, combining language education with sociocultural education, includes competencies emphasized in Japanese business culture, such as problem-solving ability, communication skills, and teamwork. This supports the necessity of a business Japanese education model that combines cultural education with practical competence, as addressed in this study, and highlights the importance of task design and teaching strategies that reflect cultural contexts to help learners adapt smoothly to actual business environments.

Furthermore, Hiramatsu (2022) regarded Japanese business emails as communicative acts and conducted a qualitative discourse analysis based on email data collected from real business settings. Hiramatsu’s research method involved building an email corpus, analyzing discourse structures according to sender–receiver relationships and situational purposes, examining the types of speech acts, and interpreting cultural and social factors, such as the Japanese business culture, hierarchy, and group harmony consciousness. From the perspective of cultural approaches, Hiramatsu emphasizes that expressions in business emails are shaped not only by business professionals’ and learners’ adherence to linguistic norms, but also by relational contexts and sociocultural expectations. This suggests the necessity of guiding learners in writing education by combining linguistic forms and cultural contexts. However, since Hiramatsu’s study did not address actual feedback processes, the current study—which analyzes the process of revising learners’ business Japanese writing—can be seen as distinct (Hiramatsu, 2022).

Recently, research on AWF systems has been gaining increasing attention in Japanese language education. A representative example is the jWriter system, which automatically analyzes learners’ sentences and supports not only vocabulary, grammar, and structure but also logicality evaluation; it does so to improve the efficiency of feedback in writing education. Lee et al. (2023) proposed methods of automatically assessing learners’ proficiency and logicality using jWriter, thereby concretely demonstrating the potential of automated writing correction in Japanese language education (Lee et al., 2023). However, research on automated correction specifically for business writing that emphasizes cultural contexts and politeness strategies remains scarce. This study is significant in that it fills this research gap and explores the potential of business writing feedback systems that are applicable in actual educational settings.

Recent studies in Japanese language education (e.g., Nicholas & Blake, 2023) further demonstrate that even advanced learners frequently experience pragmatic failure in business email writing due to limited situational sensitivity. While these studies underscore the importance of sociocultural awareness, they also reveal the limitations of instructional approaches that rely primarily on form-focused explanations or static examples. Consequently, there is a growing need for pedagogical approaches that can provide more context-sensitive and individualized guidance on the subtle sociolinguistic norms governing Japanese business communication.

3. Methods

3.1. Research Design

This study adopted a mixed-methods research design integrating quantitative and qualitative methods to examine the effects of AI-based automated feedback on Japanese business writing education. The quantitative component consisted of a structured online questionnaire administered via Google Forms using a five-point Likert scale to measure learners’ satisfaction and perceived effectiveness. The qualitative component involved semistructured interviews with a subset of consenting participants to explore their cognitive and emotional responses to automated feedback, their perceptions of its reliability, and their comparative experiences with instructor-provided feedback. This mixed-methods approach was designed to explore not only the numerical outcomes of AI-based automated feedback but also the learners’ experiential contexts and subjective acceptance in greater depth.

3.2. Participants

The participants of this study were 27 Korean undergraduate students enrolled in the Advanced Business Japanese class at Hanyang Cyber University in Seoul, Korea; they consented to participating in the Google Forms survey. Of these, 11 students participated in the online interviews. Most of the learners possessed intermediate-level Japanese proficiency. The demographic characteristics of the 27 participants are presented in

Table 1. Demographic Information of the Participant.

Characteristics		N	%
Gender	Male	7	25.9%
	Female	20	74.1%
Age	20–29	15	55.6%
	30–39	7	26%
	40–49	2	7.4%
	50–59	3	11%
Grade Level	1st	1	3.7%
	2nd	4	14.8%
	3rd	12	44.4%
	4th	10	37.1%
ChatGPT Experience	Yes	24	88.9%
	No	3	11.1%

.

3.3. Ethical Considerations

This study was conducted in accordance with established ethical standards for research involving human participants. Prior to data collection, ethical approval was obtained from the Institutional Review Board of Hanyang Cyber University (HYCU-IRB-2024-003-1).

All participants were informed of the purpose of the study, the voluntary nature of their participation, and their right to withdraw at any time without penalty. Written informed consent was obtained from all participants before data collection. To ensure confidentiality and anonymity, participants’ identities were replaced with alphanumeric codes (e.g., P1, P2), and no personally identifiable information was disclosed in the analysis or reporting of findings. All collected data were used exclusively for research purposes and were securely stored with restricted access.

3.4. Classroom Activities Using ChatGPT

This class was conducted as a 75 min online session designed to enhance learners’ practical language skills in the Japanese business environment. Learners were instructed not only in Japanese business writing, including reading, writing, grammar, and vocabulary, but also in handling specific business situations in Japan, such as first-time greetings with business partners, telephone responses, consultations, complaints, and matters related to ceremonial occasions. The class was structured to transcend vocabulary- and grammar-centered writing instruction, enabling learners to acquire both the characteristics of business communication required in real workplaces and cultural understanding at the same time. For this study, a total of four sessions were designed, as shown in

Table 2. Classroom Activities.

No.	Lesson Topic	Class Activity	ChatGPT-Based Learning Tasks
1	Basic Structure of Japanese Business Writing and Case-Specific Documents	Learning about the types of documents essential in business communication in Japan Understanding the role of Japanese business writing	Asking ChatGPT questions on topics such as “definition of business documents,” “the role of documents in Japanese business communication,” and “differences between internal (社内文書) and external documents (社外文書)” Summarizing the answers
		Learning the basic structure of Japanese business writing, focusing on the essential elements of each section (including opening text [前文], main text [主文], closing text [末文], addressee, and signature [宛て名と署名]) Learning the expressions to be avoided in Japanese business culture	Entering actual email examples from Japanese companies and asking ChatGPT to analyze whether each component (opening text [前文], main text [主文], and closing text [末文], among others) is appropriately included Generating and reviewing a checklist of taboo expressions, overly direct wording, and expressions that disregard hierarchical relationships
2	Introduction to AI-Supported Business Japanese Writing	Understanding the introduction and purpose of ChatGPT Learning about ChatGPT registration and usage, and the feedback chatbot Japanese Business Writing Tutor Practicing the activity in which the chatbot prompts are shared so that learners can create their own chatbot that can provide feedback on their writing that is tailored to their needs	Comparing business writing responses from multiple AIs (e.g., ChatGPT and Google Gemini) Evaluating performance with the same prompt Entering test sentences into the chatbot and analyzing the results Evaluating whether the expressions are overly rigid or unnatural
3	Assignment Instructions: Submission Method and Feedback Guidelines	Practicing the required formats and honorific expressions in Japanese business writing through actual document creation	Creating an assignment draft (e.g., a proposal for a new business transaction)
		Writing a request document for new transactions and appointments	Entering their draft documents into ChatGPT and requesting revisions using prompts such as “Please revise these words into more polite wording” and “Please reflect Japanese business manners in this”
4	Practical Assignment Writing and Instructor Feedback	Submitting business writing drafts Receiving joint feedback from a native Japanese instructor and a Korean instructor	Receiving draft feedback via ChatGPT and submitting links Revising and submitting drafts based on ChatGPT’s feedback Reviewing and comparing feedback from the instructor and the native Japanese instructor

, which served as the basis for analysis.

The assignment was structured around learners independently setting a specific business situation and writing an appropriate business document for that context. While completing the task, learners drafted an initial version, used ChatGPT to receive iterative feedback and revise their documents, and ultimately submitted three types of work: the initial draft of the business document, the final revised version incorporating AI feedback, and a record of multiple revisions generated through the chatbot. After submission, the instructor provided individualized feedback focusing on both the completeness of the document and the learners’ strategies for utilizing AI.

3.5. Custom-Designed Chatbot and Automated Feedback Through ChatGPT

The assignment involved the instructor designing a chatbot within ChatGPT and prompting learners to make use of it. The chatbot was created using the custom GPT function of ChatGPT, and the prompt was designed with the following roles and functions.

The role was set as a professor responsible for a Japanese business class at a cyber university, thereby maintaining the context of an actual class. The functions were to provide feedback on learners’ assignments; check the formal structure of various documents, such as requests and appointment letters; guide linguistic expressions, including honorifics; and foster an understanding of cultural contexts. The guidance emphasized maintaining conventional expressions, such as haikei (拝啓, a formal salutation) and keigu (敬具, a closing phrase denoting respect) at the beginning and end of documents, ensuring polite and clear expressions, logical development of content, and professionalism.

As shown in Figure 1, the chatbot—designed with prompts provided by the instructor—presented menu options, such as “request for document feedback,” “assistance with appointment request letters,” “review and revision of greetings,” and “checking document formats,” enabling learners to actively and easily select functions according to their needs. The chatbot was provided in conjunction with tutorial videos embedded in the class content and was integrated into activities combining AI feedback with actual document-writing tasks.

Furthermore, the chatbot feedback used in this study can be broadly divided into two categories. First, suggestive feedback was provided in the form of proposing new vocabulary or sentences that allowed learners to improve their expressions or consider alternatives. Rather than presenting the correct answer directly, this approach left room for learners to think, choose, and adjust, thereby promoting learner autonomy. Second, corrective feedback directly highlighted grammatical errors, spelling mistakes, and inappropriate vocabulary usage, offering specific revisions.

This prompt design aimed not merely to supplement linguistic aspects but to help learners recognize the triad of document organization, linguistic expression, and cultural context and to enhance their writing strategies. This represented an attempt to create an organic connection between theoretical knowledge delivery in lectures and practice-oriented learning with the chatbot.

After the automated feedback was received, a supplementary stage was implemented in which a native Japanese instructor reviewed the learners’ revised writing outcomes and provided in-depth feedback. This instructor had over ten years of experience in Japanese language education and possessed expertise not only in learners’ linguistic abilities but also in the cultural norms and social contexts of Japanese business communication. The instructor’s feedback focused on aspects that the automated feedback could not completely capture, such as the content’s logical coherence, subtle nuances of expression, and conventions of actual communication practices in Japan. While AI could highlight errors in honorifics or structural issues in formal documents, matters such as selecting a tone that conveyed genuine sincerity on an emotional level and using closing expressions that presuppose ongoing business relationships were supplemented through feedback grounded in the instructor’s cultural experience. The entire feedback process is shown in Figure 2.

3.6. Instruments

In this study, both quantitative and qualitative investigations were employed to obtain a comprehensive understanding of learners’ experiences with AI-assisted feedback. By integrating survey and interview data, the study also sought to capture the contextual and affective dimensions of learner engagement.

3.6.1. Online Survey

The questionnaire, as shown in

Table 3. Questions to Evaluate Learners’ Responses.

Seven Multiple-Choice and Eight Essay Questions	Q1. Have you ever used AI (e.g., ChatGPT) before this class? * Q1-1. If yes, for what purpose did you use it? Q2. Are you satisfied with using AI in this class? * Q2-1. Please explain the reason. Q3. Did AI help you manage your learning? * Q3-1. Please explain the reason. Q4. Did AI help you acquire knowledge of Japanese business writing? * Q4-1. Please explain the reason. Q5. Was the chatbot’s feedback helpful for your Japanese business writing? * Q5-1. Please explain the reason. Q6. Did AI help you understand Japanese business manners? * Q6-1. Please explain the reason. Q7. Did AI help you practice using language in diverse contexts? * Q7-1. Please explain the reason.
	Q8. If you have any suggestions regarding the class using AI, please describe them freely.
Multiple-Choice Question Score	Five points: Very useful Four points: Useful Three points: Moderately useful Two points: Not so useful One point: Not useful at all

Note: The asterisk (*) indicates an essay question.

, combined both open-ended and closed-ended questions and was administered via Google Forms. Learners’ overall satisfaction with the use of ChatGPT in class, the effectiveness of ChatGPT-based feedback for their writing practice, and its role in promoting the understanding of Japanese business culture were measured using seven items on a five-point Likert scale. In addition, the respondents were asked to provide open-ended answers explaining their reasons. One further open-ended question was included to elicit learners’ free suggestions.

3.6.2. In-Depth Interview

Learners’ responses to the online survey were often relatively short or fragmentary; therefore, it was difficult to explore more in-depth aspects, such as learners’ learning tendencies, their recognition and use of the class and automated feedback, and the differences in acceptance between automated feedback and instructor feedback only through the survey. Therefore, in this study, additional individual interviews were conducted.

The interviews were conducted according to the preferred time of the participants. Participants are shown in

Table 4. Demographic Information of the In-Depth Interviews’ Participants.

Learner Code	Gender	Age (Year of Birth)	Major
P1	Female	47 years (1978)	Japanese
P2	Female	35 years (1990)	Japanese
P3	Female	30 years (1995)	Japanese
P4	Female	24 years (2001)	Japanese
P5	Male	55 years (1970)	Finance, Accounting, and Taxation
P6	Female	27 years (1998)	Japanese
P7	Female	26 years (1999)	Japanese
P8	Male	27 years (1998)	Japanese
P9	Female	40 years (1985)	Japanese
P10	Female	25 years (2000)	Computer Science
P11	Female	27 years (1998)	Finance, Accounting, and Taxation

. The purpose of the research was explained to the participants, guidance on voluntary participation was provided, and consent forms for research participation were obtained before proceeding. All interviews were conducted in a one-on-one format and lasted about 30 min on average.

The interview questions were composed based on topics similar to those of the essay questions in the questionnaire, and the learners responded in more detail about their personal reasons, specific usage contexts, and perceptions of feedback. The interviews were then transcribed from the recorded content, and the authors read them repeatedly and analyzed them through discussion and review. First, open coding was conducted, focusing on the key utterances obtained from the transcripts, and these codes were categorized to examine similarities and differences among the codes.

4. Findings

This study synthesized mixed-method surveys and interviews conducted with cyber university learners to explore the impact of ChatGPT-based automated feedback on learning Japanese business writing. For complete anonymity, the learners were assigned alphabetical codes, such as A, B, and so on.

4.1. RQ1: How Does the Use of ChatGPT in Classroom Instruction Influence Learners’ Overall Satisfaction and Learning Experience?

The results of the responses to the multiple-choice questions in a survey using a five-point Likert scale are shown in

Table 5. Learner Satisfaction with Japanese Business Writing Classes Using ChatGPT.

Question	Average	Standard Deviation
Q2. Are you satisfied with using AI in this class?	4.3	0.71
Q3. Did AI help you manage your learning?	4.0	0.90

. In the descriptive answers from the Google Forms survey, many responses emphasized the convenience and immediacy of using AI, with comments such as, “It is always accessible, so learners can ask questions and receive answers whenever they need.” These responses indicate that the learners generally value AI for its efficiency in information retrieval and its ability to provide learning support without restrictions of time and place. A more detailed investigation into this aspect, based on the analysis of the interviews, is presented below.

4.1.1. Establishing Learning Plans

The analysis of the interviews revealed that many learners used ChatGPT-based automated feedback not only as a simple document correction tool but also as a means to diagnose their language proficiency and to actively plan future learning.

“It was really helpful when I wanted to figure out where my Japanese proficiency stood or when I tried to draft a learning plan. Since I’m working and have other obligations, I sometimes have gaps in my schedule, so I asked for advice on how to study during those times. I basically used it for schedule management.”

(P2)

In addition, some learners developed plans to focus on areas where they personally needed more practice. For instance, H, after repeatedly receiving corrections for errors in kanji and particle usage, concretized a plan to study those areas intensively: “Since I was often told that my kanji knowledge was lacking or that I frequently made mistakes with particles, I thought, ‘I really need to study particles more.’ So I asked, for example, ‘Could you set up an eight-week study plan for me?’ Then, a simple plan was generated. I was able to further adapt it to fit my own schedule, which was great.”

These cases demonstrate that AI-based feedback can assess a learner’s current competence and, based on this, provide a personalized learning roadmap. In this way, ChatGPT-based automated feedback shows its potential to function as a tool not only for short-term task completion but also for the creation of long-term, systematic, and self-directed learning plans.

4.1.2. Learning Records

In the survey, the learners highlighted the preservation of prior learning history as an important advantage of AI-based feedback, noting that “it can be used continuously after the initial use.” (P1) This allowed the learners to continue learning without interruption and to achieve gradual improvement based on accumulated records.

“What I liked was that there was a record of what I had previously studied, so I could continue from there.”

(P1)

Mejía et al. (2024) also reported when foreign-language learners were asked to maintain learning logs, this practice strengthened their self-directed learning attitudes and positively affected their vocabulary acquisition, expressive ability, and language fluency. The process of recording one’s learning and repeatedly reviewing it can serve as a driving force for continuous improvement in language learning (Mejía et al., 2024). Similarly, in this study, the learners’ experiences of utilizing AI-based records are significant in that they contribute to constructing individualized learning paths through goal setting and self-monitoring.

4.1.3. Reducing Learning Time

The analysis of the interviews revealed that many learners expressed satisfaction with the fact that AI tools reduced the time spent searching for materials.

“It usually takes time to check the contents of multiple websites one by one, but with AI, it was possible to collect materials more smoothly.”

(P11)

“Unlike a translator which simply provides an answer, it explains the parts that need to be corrected so that I can understand the reason, and it reduces the time spent searching, which was satisfying.”

(P2)

This tendency has also been confirmed in other studies. Boillos and Idoiaga (2025) analyzed cases of learners using AI-based writing assistance tools in university foreign language classes and reported that their accessibility to information was improved, highlighting the speed, ease, convenience, and diversity of information (Boillos & Idoiaga, 2025). Similarly, Möller et al. (2024) reported that after the introduction of an AI-based tutoring system in a university’s remote-learning environment, the time that learners spent on studying decreased by an average of 27% (Möller et al., 2024). In this way, AI-based feedback suggests the potential to improve learners’ time efficiency, thereby supporting time management and improving learning concentration in language learning environments.

4.1.4. Supplementing Existing Learning Methods with AI

In the interviews, many learners expressed the opinion that the introduction of AI did not replace traditional learning methods but rather played a role in extending and supplementing them.

“When I learned in class, I thought it was just my head. But it was easy to understand when I wrote an email myself and asked ChatGPT to fix it. When I actually used the expressions that would have passed if I had just listened to the lecture, I remembered them.”

(P4)

“Looking at only the dictionary was confusing. But ChatGPT corrected what I wrote right away, which was helpful. It felt like confirming what I had learned in class.”

(P2)

These experiences show that knowledge acquired through lectures was applied, and AI feedback supported this, thereby supplementing existing learning. In other words, within the continuous learning cycle of lectures, assignments, and AI use, the learners could strengthen learning effects by receiving real-time answers tailored to their needs through AI. This suggests that traditional learning and AI use can be connected in a complementary way to further enhance learning outcomes.

4.1.5. Expressing Concerns About Learning Dependency

During the interviews, the learners not only mentioned the advantages of AI-based feedback but also discussed its negative aspects and risks.

“In the past, I used to spend time searching and skimming through materials to get a general understanding. But now, I tend to just check the answers or results provided by AI.”

(P2)

“Since I can learn faster now, I feel like I put in less effort, and that might be a drawback. Because I put in less effort, the knowledge doesn’t stay with me for long.”

(P3)

The learners reflected on their previous learning process of searching and comparing different sources to grasp the overall context, stating that it had recently been replaced by simply checking the single answer provided by AI. They also highlighted that while knowledge acquisition has become faster with AI’s help, the reduced effort may have a negative impact on long-term retention. In other words, although AI can increase the efficiency of the learning process, it may also reduce the stage at which learners explore problems on their own, thus negatively influencing learning.

In addition, some learners mentioned dependency on prompts, noting that the quality of feedback varies, depending on how inputs are phrased, and that effective use requires learners’ ability to design questions. Considering this, instructors should seek a balance by supplementing learners’ independent inquiry activities to prevent excessive dependency.

4.1.6. Differences in Learner’s Backgrounds and Perceptions of AI Use

The learners in the cyber university environment come from diverse age groups and professional backgrounds. Many of the participants in this study already had professional experience in fields such as education, business, or public service, which led to differences in how they perceived and used AI tools during the learning process. Interestingly, it was not the learner’s age itself that influenced the perception of AI use more but professional experience and roles.

“After writing a business email with ChatGPT, I wasn’t sure whether the expressions were actually used, so I often asked people around me who knew Japanese. In the end, I only felt reassured after checking with someone with business experience; otherwise, I had to spend time searching a lot by myself.”

(P3)

“I’m more curious about how Japanese people would actually perceive my expressions than about grammatical accuracy. There may be a difference between Japanese and actual usage learned in textbooks, and I feel anxious if I don’t check the emotional reaction of the native speaker.”

(P7)

P3 and P7, who had relatively little professional experience and limited exposure to business document writing, showed a tendency to critically verify AI outputs rather than accept them at face value. In contrast, learners with extensive professional experience evaluated AI as a tool that improved the efficiency of business document writing required in the workplace and supplemented the linguistic formality necessary for professional use, drawing on their prior experience with email exchanges with Japanese companies.

“ChatGPT is meaningful for supplementing formal document expressions. I use it alongside online resources when needed to review drafts, and I judge whether to revise or accept outputs based on my own knowledge and experience.”

(P5)

P5 thus used AI selectively and actively, grounded in personal experience. This demonstrates that perceptions of AI use can vary according to professional background or occupational role, rather than simply by age.

These patterns show that background factors, such as age and professional experience, are not superficial variables but are closely related to learners’ expectations, domains of use, and perceptions of reliability. Moreover, these features did not stand out in the Google Forms survey but became more evident during in-depth interviews; indeed, the learners described their experiences and contexts in detail. This confirms that quantitative and qualitative approaches are complementary in shedding light on learners’ experiences from multiple perspectives, supporting the usefulness of a mixed-methods design in future language education research.

4.2. RQ2: How Does ChatGPT-Based Feedback Affect Japanese Business Document Writing Skills?

The results of the analysis of the responses to the survey with multiple-choice answers are shown in

Table 6. Satisfaction with Learning Japanese Business Writing.

Question	Average	Standard Deviation
Q4. Did AI help you acquire knowledge of Japanese business writing?	4.3	0.6
Q5. Was the chatbot’s feedback helpful for your Japanese business writing?	4.19	0.61

. In the open-ended responses, the learners expressed satisfaction with the immediate and convenient feedback provided. They also positively recognized the effects of correction related to grammar and expressions, such as modifying honorifics and awkward sentences into more natural forms.

4.2.1. Immediate Personalized Feedback

The learners frequently mentioned the immediacy of feedback. For instance, “Since I could receive feedback in real time, it was convenient not having to go through the instructor.” (Survey response). They also viewed positively that ChatGPT’s feedback directly suggested context-appropriate revisions or alternative expressions, enabling learners to immediately understand and apply them.

“There are times when you can’t get feedback right away. Sometimes, I study late at night, and I had to get it done right then … In that respect, being able to receive feedback immediately was convenient.”

(P2)

These interviews suggest that the learners were able to overcome time constraints during task performance and resolve difficulties encountered in writing assignments in real time, thereby sustaining learning motivation and improving outcomes. Such results align with prior research showing that immediate feedback helps resolve learner confusion, identifies learning gaps, and positively contributes to narrowing performance differences among learners (Ajogbeje, 2023).

4.2.2. Comparison of AI and Instructor Feedback

The learners mentioned the issue of the reliability of AI-based feedback. A survey respondent mentioned, “Cross-checking, which requires the user to verify the generated text, is still inevitable” (Survey response) adding that it is difficult to accept AI-generated feedback without final verification by experts. The learners, upon encountering both AI and instructor feedback, responded by comparing or contrasting the characteristics, differences, and commonalities of the two.

“Every time I made a request, I received very detailed and quick feedback. As a learner, I couldn’t confirm whether it was correct or not, but once the instructor checked it, I was able to trust the chatbot’s feedback.”

(P1)

“AI tends to give standardized answers depending on the prompt settings. But the instructor’s feedback reflected the flow of my own study.”

(P8)

AI feedback was evaluated as helpful in terms of speed and variety; however, the instructor’s feedback was perceived as fostering greater trust through in-depth, contextual advice tailored to the learning context and each individual learner. These findings are consistent with those of Asadi et al. (2025), who reported that groups receiving combined instructor and ChatGPT feedback showed greater improvement in assignment achievement, coherence, cohesion, vocabulary, and grammatical range and accuracy than groups receiving instructor feedback alone (Asadi et al., 2025). This suggests that a reliable learning experience requires a combination of an expert human’s feedback and AI-based feedback.

4.2.3. Learning Honorific Expressions and Politeness Strategies

The analysis of the interviews revealed that many learners mentioned areas of improvement in vocabulary that were appropriate to the context of Japanese business writing, the suitability of honorific expressions, and choices in sentence structure. Some learners reported that they found the chatbot’s suggestions helpful from the chatbot’s suggestions when it was difficult to judge the appropriateness of honorifics or sentence forms and that they could deepen their understanding of expressions by acquiring idiomatic phrases. Others responded that, by going through the process of revising to politer expressions, they could apply those expressions in practical business situations.

“I came to know a lot about what is called humble or honorific expressions, which I didn’t know before, and I thought it would be useful to use them when necessary.”

(P6)

“When I want to have a business conversation, if I speak in honorifics to the other person, they also respond in business honorifics, so in those cases I use it.”

(P8)

“It was very good that it told me about idiomatic expressions and why they are used.”

(P2)

These responses suggest that chatbot-based feedback contributed to strengthening learners’ sociolinguistic competence by suggesting vocabulary and honorific expressions suitable for given situations. In particular, the use of honorifics according to the level of politeness—an essential element in Japanese business writing—is difficult to completely master through textbooks alone. For example, learners mentioned that it was challenging to distinguish between sonkei-go (respectful speech) and kenjō-go (humble speech), or to choose expressions such as “I would appreciate it if you could contact me (ご連絡いただけますと幸いです)” in appropriate contexts. Repeated revision suggestions from AI feedback provide learners with multiple opportunities to learn how to use polite expressions, thereby enhancing their ability to internalize it (Ayeni et al., 2024; Maier & Klotz, 2022). This holds practical significance in that it increases the possibility of applying honorific expressions in real business contexts.

4.2.4. Reinforcement of Learning Content and Self-Reflection

The learners often used AI feedback to reconfirm and recognize content they had already learned. Rather than simply accepting feedback as corrections, they showed reflective responses, reconsidering their own writing processes and the appropriateness of the expressions they had chosen.

“When I saw GPT correcting what I had written, I thought, ‘Ah, that’s right. This is how it should be!’ and it became clearer to me … If I had just read it in a book, I probably would have overlooked it. But because ChatGPT happened to correct that sentence, I was able to clearly recognize it and felt that I could use it very effectively.”

(P9)

This process suggests that details that might have been overlooked in textbooks or lectures were clarified and that the learned content may be remembered for a longer period. In particular, personalized feedback tailored to individual learners, rather than uniform instruction, increases learner engagement and optimizes learning outcomes. As a result, AI-based feedback can be seen as contributing to a deeper understanding and stronger retention of the learned content.

4.3. RQ3: How Does Feedback from ChatGPT Affect Learners’ Understanding of Japanese Business Culture?

The analysis of the responses to the multiple-choice survey on a five-point Likert scale is shown in

Table 7. Learners’ Satisfaction with Learning Japanese Business Culture Using AI.

Question	Average	Standard Deviation
Q6. Did AI help you understand Japanese business manners?	4.37	0.67

. The results show that learners’ satisfaction with learning Japanese business culture using AI was high (M = 4.37, SD = 0.67), indicating that most participants perceived AI feedback as helpful for understanding Japanese business manners.

In the open-ended responses, it was evident that the learners reacted positively to AI feedback in terms of understanding cultural contexts. One learner answered, “By asking it to set up a business situation, I was able to practice language use in context” (Survey response), referring to the effect of linking language learning to business manners and conventions. Another response stated, “Since I have no prior experience with business in Japan, using AI made it easier to establish guidelines” (Survey response), which suggests that AI contributed to helping learners with limited real-world experience accumulate indirect experiences to compensate for cultural differences.

4.3.1. Learning Business Situation-Specific Communication Techniques

In the individual interviews, the learners mentioned that they received support in developing the communication skills required in specific business situations, including telephone etiquette, business card exchange, and email writing. In particular, the key learning elements identified were the selection of the initial utterance appropriate to the situation; linguistic adjustments according to the interlocutor’s age, status, or gender; and the consideration of the impression and emotional response conveyed by expressions.

“When I make a phone call, I find it really difficult to know what I should say first.”

(P2)

“It [the appropriate way of communicating in business situations] may depend on the language being learned. But depending on whom you are dealing with—such as the person’s gender or age—it is better to receive information and engage in practice tailored to these specific contexts.”

(P5)

In actual business settings, language choices vary depending on the situation and the interlocutor’s status, age, and relationship; therefore, cultural context must also be considered. Interaction with ChatGPT fills this gap by providing situation-specific expressions tailored to business contexts and by checking conventions embedded in Japanese business culture in real time, thereby distinguishing itself from textbook-centered learning. These results suggest that AI feedback can contribute to expanding learners’ intercultural communicative competence.

4.3.2. Learning Expressions by Understanding Regional Culture

The learners mentioned that they could acquire cultural nuances and contexts, which are difficult to grasp without being a local, by using AI in a relatively easy manner.

“When a Japanese person says that people in Tokyo don’t do it this way but people in Kansai do, I can’t really know that since I’m not a local. Looking it up in a dictionary doesn’t reveal such differences in perception. For that kind of thing, I thought it was better to use AI.”

(P9)

“Nuances that only someone who has experienced them directly can feel don’t appear in books, so I often ask about them on YouTube or with GPT.”

(P9)

“I searched how people behave at drinking parties when meeting with clients.”

(P8)

These findings align with those of Klimova and Chen (2024), who reported that, through virtual simulations, AI-based learning tools can guide learners into the ethnorelative stage of cognitive and cultural acceptance (Klimova & Chen, 2024). The findings of this study likewise suggest that ChatGPT-based feedback can function as a learning resource that enhances learners’ ability to contextually understand and express regional characteristics and cultural conventions. This highlights the potential of AI to complement cultural sensitivity, which may be insufficient in textbook-centered, standardized learning environments.

5. Discussion

This study explored the learners’ experiences in writing Japanese business documents from three perspectives: learner satisfaction, feedback effectiveness, and cultural understanding. Synthesizing these findings is meaningful for clarifying how ChatGPT is received, depending on individual learners’ strategies, and elucidating what educational implications it holds as an integrated model for language and culture learning.

First, this study reported that the learners generally expressed high satisfaction with using ChatGPT in the class, largely because it provided substantial support for study planning, record keeping, and time saving. The learners reported that they went beyond merely completing assignments, actively using AI to articulate their learning goals and to devise step-by-step plans and thereby fostering self-directed learning. The accumulated interaction logs with ChatGPT functioned as learning records, giving learners opportunities to reflect on and monitor their language use. Compared with traditional lecture notes, this process allowed for more precise reflection and continuous improvement of learning trajectories. The learners also repeatedly emphasized that ChatGPT reduced the time required for their information search and resource verification—an especially meaningful benefit for adult learners facing significant time constraints. Moreover, they perceived ChatGPT as complementing, rather than replacing, traditional textbooks and dictionaries. At the same time, some learners voiced concerns that deepening reliance on AI may weaken their critical thinking and problem-solving skills. Thus, while ChatGPT use enhances learners’ efficiency and engagement, its implementation must be carefully designed to ensure that autonomy and critical reflection are maintained.

These results align with the findings of Jin et al. (2023), who reported that AI applications support learners in planning and self-regulation. Building on this, the present study further highlights that adult learners in an online university particularly valued ChatGPT as a tool for overcoming time limitations and ensuring learning continuity. At the same time, this study highlights that adult learners in an online university particularly valued ChatGPT as a tool for overcoming time limitations and ensuring learning continuity. This sets them apart from traditional undergraduate populations. For educators, this suggests the need to design tailored strategies that accommodate the realities of adult learners who balance study with work and family responsibilities. Furthermore, because learners’ perceptions of AI use varied significantly, depending on their professional experience, future instructional design should include differentiated guidance according to the learners’ experience levels.

Second, this study confirmed that through the immediate and personalized feedback provided by ChatGPT, the learners could promptly correct errors and effectively perform repeated practice. This effectively supplemented the physical constraints of textbook-centered learning or instructor-centered classes while providing the learners with individualized learning paths, thereby contributing to enhanced learning immersion and self-directedness. Liu’s (2024) meta-analysis emphasized that automated feedback supports repetitive writing training, reduces the burden on instructors, and has a positive effect on improving learners’ writing outcomes (Liu, 2024). The results of this study can be regarded as an extension of such previous research to the specific learning context of Japanese business writing.

In particular, one of the aspects that the learners emphasized was honorifics and polite expressions. This can be understood as the process of acquiring the formal conventions required in Japanese business discourse and the characteristics of business documents. This study is significant in that it moves beyond the conventional focus on grammatical accuracy in AI feedback and expands the perspective to include sociolinguistic dimensions, such as honorifics and politeness.

These results provide clear implications at the level of instructional design. Instructors need to conduct preclass prompt literacy training, such as sharing self-monitoring checklists, so that learners can effectively compose prompts. Learners must internalize a dialogic, critical verification process by asking better questions that allow for revisions and counterarguments regarding AI feedback results. Linguistic elements, such as honorifics and polite expressions, can achieve maximum learning effectiveness when instructor feedback is provided alongside, for instance, by selecting and reviewing parts of the dialog logs between AI and learners. This reconfirms, as Lan and Zhou (2025) highlight, the importance of designing feedback so that learners can actively accept and utilize it (Lan & Zhou, 2025). Therefore, in future instructional design, it will be necessary to guide learners not to rely entirely on AI but to use AI feedback as a strategic learning resource. Through such balanced utilization, AI-based feedback can be positioned as an educational partner that strengthens learners’ active participation, self-directedness, and strategic learning competence.

Third, this study demonstrates that ChatGPT-based feedback can play a meaningful role in expanding learners’ understanding of Japanese business culture. Learners indirectly experience business situations, such as the use of telephone etiquette and business card exchange, thereby engaging in practical communication that is not easily addressed in textbooks and dictionaries. This highlights the significance of showing that interaction with ChatGPT can enhance workplace readiness, making it applicable to real-world practice.

In addition, the learners used ChatGPT to quickly explore and compare regional language habits and cultural conventions. This indicates that AI can function as a tool to foster not only linguistic ability but also cultural sensitivity, thereby contributing to the development of international communication competence, which aligns with previous studies (Karakas, 2023). However, some prior research has highlighted that while AI performs well in explicit speech acts, it does not effectively handle indirectness, ambiguity, and cultural variability (Eragamreddy, 2025). This study confirmed that the instructor’s cultural interpretation and supplementation remain important.

Therefore, future instructional design should systematically combine the instructor’s explanations that reinforce cultural contexts with case-based learning that allows learners to simulate actual business scenarios. Instructors should also provide diverse scenarios that could be culturally misinterpreted, enabling learners to elicit more precise cultural information and develop an attitude that considers cultural outcomes. Through such methods, it is necessary to establish strategic mechanisms that encourage learners to consider cultural contexts on their own and to critically verify ChatGPT-based feedback.

Unlike the majority of learners who responded positively, some noted issues, such as the mechanical nature of chatbot feedback or the technical difficulties that older learners face in accessing ChatGPT. The following comments reveal that these challenges stem not merely from the quality of the AI tool itself but from gaps in digital literacy needed to use AI effectively: “It would be better if there were guidelines for entering prompts” (P10), and “since learners must ask questions themselves to obtain the answers they want, a certain degree of creativity is also required” (Survey response). This suggests that ChatGPT-based learning provides learners with new opportunities while simultaneously imposing new learning tasks, such as developing prompt formulation skills. This analysis indicates that AI feedback is not simply a tool that delivers information unilaterally but rather an interactive resource whose effectiveness can vary, depending on the user’s competence, attitude, and technological environment. Accordingly, future instructional design requires supplementary measures that consider learners’ age and digital experience, including prompt-writing training, usage guidelines, and step-by-step support.

6. Conclusions

This study addressed the question of how ChatGPT-based automated feedback affects learners’ Japanese business writing and their understanding of Japanese business culture. By exploring this relationship, the study explored the impact of ChatGPT-based automated feedback on Japanese business document learning, demonstrating that AI can function as an educational partner that goes beyond mere linguistic correction to integrally connect learners’ linguistic and cultural learning. Particularly within the context of adult learning at a cyber university, it is noteworthy that learners of diverse ages and professional backgrounds experienced self-monitoring, the reorganization of learning strategies, and the cultivation of cultural sensitivity while using AI. Taken together, the findings indicate that AI-based feedback can play a transformative role in language education by fostering both linguistic accuracy and sociocultural competence. Furthermore, this study suggests that hybrid feedback—grounded in the interaction between AI, instructors, and learners—may emerge as a new paradigm in foreign language education.

However, this study has several limitations. First, it is restricted to Korean learners as its research subject. Since Japanese and Korean are typologically related languages, these findings cannot be generalized to other languages. More empirical learning cases across diverse languages need to be examined. Second, this study is limited to online learners. Because all participants studied in an online learning environment, which differs from face-to-face classes in social interaction and feedback dynamics, the findings may not fully generalize to traditional university settings. Third, in this study, ChatGPT was optimized to specific prompt settings and assignment contexts, meaning that, even with the same AI tool, different results could arise depending on prompt design and how learners employ the tool. This implies that the effectiveness of AI-based feedback is influenced by learners’ prompt literacy as well as the complementary role of instructors.

This study offers practical implications for language educators in online higher education. The findings suggest that AI-based feedback is most effective when implemented through a hybrid model that combines immediate AI-generated suggestions with instructor guidance, particularly for cultural and contextual refinement. In addition, instructional design should include training in AI prompt literacy and critical evaluation to ensure that learners use AI strategically and reflectively. These findings underscore the importance of pedagogically grounded, culturally responsive AI integration in business language education.