This research explores the design and efficacy of a personalized AI chatbot, developed using the GPT-4.0 API,

to support seventh-grade students at Kinderfield-Highfield School Pontianak in mastering English process

student checklists following implementation. Results demonstrated that the chatbot was highly effective in

boosting students' motivation, engagement, and writing effectiveness, particularly during the drafting and

editing phases. However, the study also identified key areas for refinement, such as enhancing the clarity of

feedback and fostering greater independent thought during the pre-writing and revising stages. Experts

affirmed the tool's convenience, educational value, and potential for future use. The study concludes that while

AI chatbots present a transformative tool for personalized writing instruction, their design must strategically

encourage critical thinking rather than dependence to fully realize their benefits in the L2 classroom.

The integration of Artificial Intelligence (AI) in education has revolutionized traditional teaching

teachers’ instruction. The development of social media platforms enables them to discover and interact with AI

concepts and tools from an early age. As a result, students are increasingly seeking ways to incorporate AI into

their learning processes, including writing homework from teachers at home. Since the ethical guidance for

students in school institutions is unavailable, students may become overly dependent on these tools, using

them as shortcuts rather than developing their own critical thinking and problem-solving skills. The easy

access to AI-driven platforms may discourage them from investing the necessary time and effort to thoroughly

understand the material.

The process-oriented writing approach demonstrates effective writing, as it requires an understanding of

learners' cognitive processes (Williams & Burden, 1997). Instead of concentrating only on the finished

product, this method emphasizes the nature of writing (Hyland, 2003). Research has shown that process-

students to recognize and internalize the key components of writing processes (Hyland, 2003; Yang, 2018).

These roles that teachers are expected to fulfil can create significant cognitive and emotional demands, which

can make it difficult for them to manage the varying needs of individual students effectively. Teachers are not

only responsible for delivering content but also for acting as mentors, motivators, and emotional supporters, all

while addressing the unique challenges and learning styles of each student (Lipson et al, 2000).

Hence, AI technologies offer significant integration in the English writing process. Every stage, such as pre-

writing, drafting, revising and editing, requires writers with a varied range of specialized skills (De Larios et

al., 2002). Generative AI can adapt to these multiple stages and capable to understand task contexts and user

intents (Zou & Huang, 2023). These technologies can enhance the specific needs of each stage, providing

personalised assistance that aligns with the writer's progress, such as suggesting ideas during brainstorming,

offering structural guidance during drafting, or refining language use in the revision process. This ability to

provide stage-relevant help can elevate the writing process, making it more efficient and accessible for L2

text raises significant ethical concerns, such as plagiarism and academic dishonesty, as learners may present

AI-produced content as their own work (Yuan et al., 2024). These challenges involve careful consideration of

how AI is integrated into L2 writing education, ensuring that it complements, rather than replaces, the learning

process while upholding academic integrity.

To leverage the opportunities of generative AI while strategically addressing its challenges, it is essential to

consider technological advancements and instructional design in tandem and create innovative environments

for enhanced language learning. As one such approach, this paper explored the development of personalised AI

Chatbots with ChatGPT, especially GPT-4.0 API, as the brain. The researcher used prompt fine-tuning

techniques to develop GPT-4.0 API-based chatbots as personalised writing tutors that offer stage-specific

feedback, prevent unquestioning reliance on AI, and promote writing development. Lastly, the researcher also

The study employed a developmental design. According to Seels and Richey (1994), a developmental study is

a systematic investigation of designing, developing, and evaluating instructional programs, processes, and

products that are evaluated against the criteria of internal consistency and effectiveness. During the

development of AI Chatbots, the researcher employed the ADDIE model by Branch (2009). The ADDIE model

The data collection for this study involves a comprehensive approach to understanding the needs of seventh-

grade students and evaluating the effectiveness of AI chatbots in addressing those needs. The technique

employed includes distributing a needs analysis questionnaire, an expert evaluation questionnaire, and a

student checklist. A needs analysis is conducted before designing the AI chatbots to gather data on the target

needs and learning needs of the students. This data is essential to address the specific challenges faced by

learners, such as improving motivation and overcoming anxiety, particularly in areas like time management,

grammatical accuracy, and fluency, as observed in the English subject class during writing assignments and

English language information about the learners (their current skills and language use) The questionnaires also

regarded as one of the ways to evaluate products from the participants involved.

Once the AI chatbot prototype was developed, an expert evaluation questionnaire was given to evaluate the

content and design appropriateness of the chatbot. This process ensures that the tool is effective, engaging, and

aligned with educational goals. Lastly, after students had interacted with the AI chatbots, a student's checklist

was distributed to gather their opinions on the chatbot's functionality and usefulness in supporting their writing

development. These data collection methods provide a well-rounded understanding of the students' needs, the

chatbot's effectiveness, and areas for further refinement.

The data came from the Needs Analysis Questionnaire, the Expert Evaluation Questionnaire, and the Student

Checklist. To analyse the students' needs, the researcher examined the data in the form of questionnaires. The

whether its design is user-friendly and effective. Responses to these statements were collected on a Likert

scale, with scores ranging from 1 to 4. The scale ranged from "Strongly Agree" (SA) with four points, "Agree"

(A) with three points, "Disagree" (D) with two points, to "Strongly Disagree" (SD) with one point. Referring to

Kang and Sung (2024), the Likert scale questions about the effectiveness of the chatbots were organized into

five themes: entertainment, convenience, writing effectiveness, interest in writing, and future use. These

themes were applied to both the overall chatbot experience and each of the four individual chatbots. The

chatbot, providing valuable data for further improvements.

Meanwhile, the Student Checklist was used to gather feedback from the students after they had used the AI

whether AI tools would be effective in improving their writing experience. There were 25 students participated

to engage with English writing more frequently. A significant number of students (10 out of 25) have been

learning English for less than a year, which could indicate they are still developing foundational writing skills.

The remaining students have varying levels of experience, with a few having learned English for several years.

Over half of the students use technology to assist in their writing, indicating that digital tools could play an

essential role in improving their writing skills. This suggests that integrating a chatbot into their learning

process could be beneficial. The majority of students report that their writing skills are poor or need

improvement. This highlights the need for support, particularly in areas such as grammar, organization, and

revision. The most common challenges are related to generating ideas and writing clear, complete sentences.

These are areas that a chatbot could address by offering brainstorming help, sentence structuring, and grammar

assistance. Feedback is not always frequent, with a few students receiving little to no feedback. This suggests

the need for more regular feedback mechanisms, which could be supported by a chatbot that provides instant

process becomes more efficient and less time-consuming for students. Each stage has the same flow. What

differ each stage is the prompts which will be in the next part of the topic. The following is the workflow of

Pre-Writing stage:

Revising: You are an academic writing facilitator assisting students in the revising stage with roles and

independent error correction by the students. In order to give better feedback, you have to help students correct

errors in grammar, spelling, and punctuation in their written work. Do not rewrite the text for them; instead,

give indirect feedback to help students correct the error themselves. Figure 3 shows where the placement of

every fine-tuned prompt in N8N:

the drafting stage, the criteria included the extent to which the brainstorming results from the pre-writing stage

were reflected in the draft and the comprehensibility of the draft. For the revising stage, the criteria were the

degree of content revision and the degree of organizational and structural revision. Finally, in the editing stage,

the criteria were the completeness and readability of paragraphs as well as the level of editing, particularly

regarding errors in grammar, spelling, and punctuation.

improve writing skills. Entertainment value had mixed responses, with one IT developer questioning its

helped them generate ideas. However, there were notable discrepancies in terms of independence during this

stage. While some students felt they could generate ideas on their own with the chatbot’s help, others felt the

need for more structured guidance. This suggests that the chatbot could improve by offering more tailored

prompts or strategies to encourage independent thinking. The Drafting stage showed a more consistent level of

positive feedback regarding the chatbot’s help in organizing ideas and providing clarity in sentence structure.

clarity. Enhancing the feedback on punctuation and offering faster feedback could make the editing process

even more effective. The chatbot was generally effective in supporting students through all stages of the

writing process. However, there are areas for improvement, particularly in independence during pre-writing,

clarity in revisions, and editing speed. With these improvements, the chatbot could become an even more

powerful tool for enhancing students' writing skills across the board.

The AI chatbot designed to assist students in process-oriented writing has shown promising results. It offers

support in idea generation, organization, content development, and error correction, making the writing process

more accessible and efficient. The chatbot was generally effective in motivating students and providing

feedback, improving their writing skills in grammar, sentence structure, and clarity. Furthermore, experts

Macmillan.

Korean English Education Society 12(1), 123-139.

S. Ransdell and S. Barbier, eds., New Directions for Research in L2 Writing, 11-47. Springer.

study of an L2 writer. Assessing Writing 30, 3-20.

Mirae English Language and Literature 19(4), 365-390.

Journal 101(2), 209-231.

31(2), 200-210.

writing and their effects on performance. Journal of Second Language Writing 24(1), 1-19.

Advanced Multidisciplinary International Journal 2(SI), 32-37.

perspectives. The Journal of Linguistic Science 82, 117-136.

Washington DC: Association for Educational Communications and Technology.

doctoral students. Education and Information Technologies 29(11), 13201-13219.