INTERNATIONAL JOURNAL OF EDUCATION AND HUMANITIES
e-ISSN: 2829-8675 .
-ISSN: 2830-4578 Volume.
Issue.
November .
: 191-202 DOI: https://doi.
org/10.
56314/ijoleh.
The Integration of Artificial Intelligence (AI): ChatGPT in Enhancing StudentsAo Speaking Ability Nurmansyah*1.
Nur Ina Syam2.
Fahmi Room3 1,2,3University of Muhammadiyah Bulukumba.
Bulukumba City.
Indonesia Correspondence* E-mail:
nurmanbaek4@gmail.
Received : 20 October 2025 Accepted : 10 November 2025 Published : 11 November 2025 Copyright .
2025 Author.
Nurmansyah.
Nur Ina Syam.
Fahmi Room This work is licensed under a Creative Commons Attribution-ShareAlike 4.
International License.
AbstractA This study investigated the integration of Artificial Intelligence (AI), particularly ChatGPT version 4o, in improving studentsAo speaking ability at the second grade of SMKN 1 Bulukumba.
Employing a quasiexperimental design, 34 students were divided into an experimental class and a control class, each consisting of 17 students.
The experimental class received speaking instruction through a blended learning model integrating ChatGPT, while the control class was taught using traditional methods.
Data were collected using pre-tests and post-tests measuring five components of speaking: pronunciation, grammar, fluency, vocabulary, and comprehension.
Findings revealed that the experimental class demonstrated a significant improvement, with average scores increasing from 70 to 75.
65, while the control class showed minimal change .
rom 67 to 67.
Statistical analysis using an independent samples t-test showed a significant difference .
= 0.
, and the effect size (CohenAos d = 1.
indicated a strong effect.
These findings suggest that integrating ChatGPT enhances studentsAo speaking ability, providing personalized and interactive learning experiences that traditional methods often lack.
The study emphasizes the potential of AI to support language learning, particularly in improving studentsAo grammar, fluency, vocabulary, pronunciation, and comprehension.
Keywords:
Artificial Intelligence.
Speaking Ability Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
ChatGPT, .
Page INTRODUCTION In the past few years, the rapid growth of Artificial Intelligence (AI) has transformed various aspects of human life, including education (Lambert.
, & Stevens.
M, 2.
Among the many uses of AI in education, one especially promising field is its integration with language learning platforms (Rusmiyanto et al.
, 2.
With English being the globalized world's lingua franca, the efficiency of English language learning tools is essential for those looking to enhance their communication and extend their chances in education and the workplace.
Since methods of instruction remain monotonous and often rely on textbooks.
English language learning, especially in learning speaking is perceived to be less successful and frequently encounters obstacles such as limited participation, lack of customization, and lack of opportunities to practice speaking (Saleem Khasawneh, 2023.
Suban, 2.
As we know, that speaking ability and proficiency is a vital aspect of second language acquisition and it stands out among the four language ability.
listening, speaking, reading, and writing, as it directly reflects an individualAos capacity to express thoughts, ideas, and emotions in real-time interaction (Andrin et al.
, 2.
In contrast.
AI-powered language learning platforms have the potential to overcome these deficiencies by providing personalized, interactive, and adaptable learning experiences that are suited to each student's specific needs (Pikhart, 2.
One highly effective application of AI in language processing, often referred to as chatbots and intelligent virtual assistant, is Chat Generative Pretrained, also known as ChatGPT (Deng, 2.
ChatGPT, an advanced chatbot technology created by OpenAI, allows users to steer conversations according to their desired length, format, style, language complexity, and level of detail.
This new AI tool was introduced to the public on 30th November 2022, and quickly gained over one million subscribers in its first week (Fahada et al.
, 2.
As OpenAI explained on the ChatGPT introduction page, they trained this model using Reinforcement Learning from Human Feedback (RLHF), similar to the method used for InstructGPT but with slight differences in the data collection process.
The initial model was trained with supervised fine-tuning, where human AI trainers generated conversations, acting as both the user and the AI assistant (Songsiengchai et al.
, 2.
These trainers used model-generated suggestions to help formulate their responses.
The new dialogue dataset was then combined with the InstructGPT dataset and converted into a dialogue format.
Until now, the updates to ChatGPT have reached ChatGPT-4o, an improvement from the previous version.
ChatGPT-3.
5, and there are likely to be further updates in the future.
ChatGPT-4o enables users to engage in reciprocal conversations not only through text but also with voice, which will respond in real-time.
The rapid adoption and continuous advancements of ChatGPT underscore as a transformative tool in language learning.
Utilizing AI, especially models like ChatGPT, educator can provide more engaging and motivating learning experiences (Almulla, 2.
Its capacity to understand and generate human-like text makes it an ideal assistant for language students, providing the immediate Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page feedback and customized interactions that traditional textbooks and method often fail to provide (Songsiengchai et al.
, 2.
This is evidenced by several teachers who primarily depend on textbooks without considering their studentsAo specific needs.
A similar situation occurs at SMKN 1 Bulukumba, a vocational high school in Bulukumba, where English instruction is largely textbooks-based.
This approach limits studentsAo vocabulary, leads to inadequate pronunciation and grammar, and results in their speaking abilities lagging significantly.
Given the advantages of SMKN 1 BulukumbaAos location in the city center and its access to adequate technological resources, there are many chances to leverage ChatGPTAos capabilities to enhance language teaching.
By integrating this AI tool, they aim to address the challenges mentioned above, as it can effectively support areas such as vocabulary building, pronunciation, grammar, and overall speaking Despite its benefits incorporating ChatGPT into language learning is not without A significant issue is the dependence of technology, which might be inaccessible to some students due to economic or infrastructural constraints.
Furthermore, the success of AI in language learning largely hinges on the quality of the model and its training data (Tiwari et al.
, 2.
Inaccurate or biased response from the AI could mislead students and perpetuate incorrect language use.
To address the existing gap in language learning effectiveness, particularly in speaking it is essential to integrate traditional methods with AI-based tools.
Traditional methods provide a solid foundation for language education, while AI tools like ChatGPT can enhance this foundation with personalization, interactivity, and real-time feedback.
Both educators and technology have distinct yet complementary roles that can significantly enhance English proficiency among Educators play a crucial role in guiding students through their language learning journey.
They bring a human touch, cultural context, and pedagogical expertise that AI cannot fully replace.
Educators can use AI tools to supplement their instruction, providing students with additional practice and feedback outside the classroom.
Given the topic's relevance and the evolving landscape of language learning, coupled with the scarcity of recent research utilizing ChatGPT4o as a medium, the researcher is motivated to explore this area.
Literature Review Speaking in Language Learning Speaking is a productive skill that involves expressing meaning through verbal Speaking is a dynamic process of constructing meaning through the production, reception, and processing of information (Nurjannah, 2022.
ran et al.
, 2.
It encompasses both interaction and performance, requiring not only linguistic accuracy but also pragmatic competence.
(Alphariyadi & Dewi, 2.
added that speaking is defined as a means of expressing or communicating opinions, feelings, ideas, and more through verbal interaction.
It is also described as an interactive process of constructing meaning, which involves producing, receiving, and processing information.
It can be concluded that speaking refers to Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page a productive ability where speakers must articulate words or sentences with specific meanings that listeners can receive and understand.
The major components of speakingAipronunciation, grammar, vocabulary, fluency, and comprehensionAiserve as the foundation for oral communication (Harris, 1.
Effective speaking instruction should thus emphasize not only linguistic form but also interactional function, allowing learners to use language meaningfully and Artificial Intelligence and ChatGPT in Education Artificial Intelligence (AI) enables computers to simulate human-like reasoning and language comprehension.
Its subfieldsAicomputer vision, speech-to-text, and natural language processingAiallow the creation of systems capable of interpreting and generating natural language (Chiu et al.
, 2.
ChatGPT, a product of OpenAI, is a conversational AI model based on the Transformer architecture.
The modelAos latest version.
GPT-4o, supports multimodal input and output .
ext, audio, and imag.
, allowing near-human response times in conversation (OpenAI, 2.
This innovation has sparked significant interest in education due to its ability to act as a conversational partner, tutor, and feedback provider.
Prior research (Jufriadi, 2024.
Muniandy & Selvanathan, 2024.
Zakiyah, 2.
has shown that ChatGPT can significantly improve studentsAo speaking proficiency, particularly in fluency, grammar, and vocabulary.
However, few studies have explored the integration of ChatGPT-4o, which includes voice-based interaction, offering more authentic communication practice.
This research addresses that gap by evaluating its impact on vocational high school studentsAo speaking ability METHOD This study employed a quasi-experimental design with a non-equivalent control Two classes were selected: an experimental class that received ChatGPTintegrated instruction and a control class taught through traditional methods.
Both groups completed pre-tests and post-tests to measure their speaking ability.
The population consisted of 325 second-grade students at SMKN 1 Bulukumba, from which 34 students were randomly selectedAi17 from XI Akuntansi II as the experimental group and 17 from XI NKPI as the control group.
Data were collected through speaking tests assessing pronunciation, grammar, fluency, vocabulary, and comprehension, adapted from (Harris, 1.
SPSS 23 software was used to analyze statistical results, including normality, homogeneity, t-tests, and effect size (CohenAos During the second semester of the 2024Ae2025 school year, the treatment took place over five talks, with each meeting lasting about 90 minutes.
It was a blended learning method that used both traditional teaching and ChatGPT-4o as a conversational partner.
The control class, on the other hand, got traditional, teacher-centered teaching without any AI tools.
The goals and topics of both Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page groups were the same, but the ways they learned were different because of ChatGPT.
Students learned about the goal of the study and how to use ChatGPT to improve their English speaking skills, with a focus on the topic "Giving Opinions and Responses.
" They used sample prompts to try out voice and text functions to get used to them and lower their tech anxiety.
The second meeting helped students improve their everyday speaking skills.
After getting clear instructions on how to say what they thought, students practiced using ChatGPT, getting immediate feedback and having conversations about language and idiomatic usage.
The third meeting focused on correct language, especially the use of passive voice.
ChatGPT gave changes and explanations in real time, which helped with personalized practice and getting feedback right away.
At the fourth meeting, students learned how to use conditional clauses to build complex sentences.
They used ChatGPT to mimic real conversations and get feedback on their pronunciation through its voice feature.
In the fifth session, students worked on their communicative production through role-playing activities.
They used ChatGPT to build and improve dialogues that they would then act out in During the treatment, the researcher served as a guide, making sure that the ChatGPT feedback was interpreted correctly and keeping the focus on In comparison, the control class listened to lectures and did pair speaking activities using textbooks.
They did not have access to digital interaction or feedback from AI.
RESULTS AND DISCUSSION
Descriptive Result The study aimed to determine whether integrating ChatGPT-4o into speaking instruction could significantly enhance studentsAo English speaking ability.
Both the experimental and control classes were evaluated through pre-tests and post-tests, assessing five core components: pronunciation, grammar, fluency, vocabulary, and In the pre-test, the experimental class achieved an average score of 70.
00, while the control class obtained 67.
These results indicate that both groups began at a relatively similar proficiency level, with only a slight difference of three points.
After the five-meeting treatment, the post-test results showed a substantial improvement in the experimental class, which reached an average of 75.
65, whereas the control class achieved only 67.
This difference highlights the positive impact of ChatGPT integration on studentsAo speaking performance.
When analyzed by speaking components, the experimental class demonstrated notable improvements across all five areas.
Grammar scores increased by 9.
pronunciation by 8.
4%, comprehension by 8.
2%, fluency by 8.
0%, and vocabulary These results suggest that the AI-supported instruction did not benefit only one aspect of speaking but enhanced studentsAo overall communicative In contrast, the control group displayed minimal progress in all components, suggesting that traditional instruction alone did not produce significant development within the same time frame.
Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page Statistical Analysis Table 1.
Independent Samples Test for Pre-test Assumption Equal variances Equal variances not Sig.
Sig.
-taile.
Table 2.
Independent Samples Test for Post-test Assumption Equal variances Equal variances not Sig.
Sig.
-taile.
Note.
*p O .
An independent samples t-test was conducted to determine whether there was a statistically significant difference between the speaking ability scores of students in the experimental and control groups, both before and after the treatment.
The analysis was performed using SPSS version 23 Table 1 presents the results of the t-test for the pre-test.
The LeveneAos Test for Equality of Variances shows an F value of 0.
085 with a p-value of 0.
772, which is greater than This indicates that the assumption of equal variances was met, allowing the use of the first line of the t-test results.
The t value obtained was 1.
162 with df = 32 and a Sig.
-taile.
value of 0.
Since the significance level exceeded 0.
05, it can be concluded that there was no significant difference in the speaking ability of the experimental and control groups before the treatment.
This finding confirms that both groups were homogeneous at the outset of the study, which validates the fairness of the experimental design.
Table 2 shows the t-test results for the post-test.
The LeveneAos Test yielded an F value 260 and a p-value of 0.
614, again indicating homogeneity of variance between the two groups.
The t value was 3.
647 with df = 32 and a Sig.
-taile.
value of 0.
Because the significance value was less than 0.
05, the result indicates a statistically significant difference between the experimental and control groups after the treatment.
This means that the students who were taught using ChatGPT-4o performed significantly better in their post-test compared to those who were taught through traditional instruction.
The finding provides strong evidence that the integration of ChatGPT into speaking instruction had a positive effect on studentsAo learning Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page Table 3.
Group Statistic Class Experimental Controlled Table 3 displays the descriptive statistics for the post-test scores of both the experimental and control groups.
The experimental group, which was taught using ChatGPT as a learning tool to enhance speaking ability, consisted of 17 students with a mean score of 75.
65, a standard deviation of 6.
547, and a standard error mean of 1.
On the other hand, the control group, which was taught using conventional methods, consisted of 17 students with a mean score of 67.
76, a standard deviation of 6.
047, and a standard error mean of 1.
To further determine the magnitude of the difference, the effect size was calculated using CohenAos d formula.
The obtained value of d = 1.
25 indicates a large effect size, based on CohenAos .
benchmark where 0.
2 represents a small effect, 0.
5 a medium effect, and 0.
8 or higher a large effect.
This large effect size demonstrates that the treatment was not only statistically significant but also educationally meaningful, signifying a substantial improvement in studentsAo speaking ability as a result of ChatGPT integration.
Discussion The result of this research indicates that the integration of ChatGPT into the learning process was effective in enhancing the speaking abilities of eleventh-grade students at SMKN 1 Bulukumba.
Moreover, the outcomes of this research support and reinforce previous research conducted by (Jufriadi, 2.
and (Zakiyah, 2.
both of which found that the use ChatGPT in the learning process positively impacts studentsAo speaking abilities.
Based on the data, it is evident that during the pre-test, there was a noticeable difference between the two classes, with the average scores being 70 for the experimental class and 67 for the controlled class.
In contrast, the post-test results showed that while the controlled class maintained an average score of 67.
76, the experimental class experienced a significant enhancement, reaching an average The researcher found that students in the experimental class responded differently to the learning process with ChatGPT compared to those in the controlled class when the study was first introduced.
Students in the experimental class appeared more engaged and enthusiastic, whereas students in the controlled class showed more neutral reaction, as their learning approach relied solely on group discussions and individual work without the use of ChatGPT.
This suggests a difference in students' motivation influenced by the learning approach.
Throughout the treatment process, the researcher observed that students in the experimental class showed increased enthusiasm.
Several students made strong efforts to practice speaking with ChatGPT, and the researcher consistently Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page encouraged them to ask for help whenever they faced difficulties.
As a result, many students in the experimental class showed greater enhancement and achieved higher scores compared to those in the controlled class.
This is evident from the gained scores, where the highest increase in the experimental class was 14 points .
rom 65 to .
, while in the controlled class, the highest increase was only 3 points .
rom 64 to .
When analyzing the speaking components, clear differences were observed between the experimental and control groups.
The students in the experimental group, who used ChatGPT via their smartphones, showed noticeable enhancement in their speaking level, particularly in terms of pronunciation, vocabulary usage, and grammatical accuracy.
Their responses during speaking tasks became more structured and confident, with a wider range of vocabulary and more accurate sentence constructions.
This suggests that the interaction with ChatGPT served as a valuable resource for practicing spoken English in a flexible and personalized manner.
In contrast, the control group, which learned through traditional instruction without AI assistance, showed less progress.
Many of these students continued to struggle with repetitive word use, limited vocabulary, and grammatical mistakes.
These difficulties often disrupted their speech flow, making it harder for them to express ideas clearly and effectively.
One of the most significant observations was the difference in sentence complexity between the two groups.
The experimental group frequently constructed more complex and coherent sentences, while students in the control group tended to produce simpler or fragmented speech.
This difference suggests that regular practice with ChatGPTAos interactive responses helped students internalize correct sentence patterns and apply them in spoken Unlike in previous studies where students might have struggled with technology adoption, in this research, all students were already familiar with using ChatGPT on their smartphones, which made the implementation smooth and efficient (Fan et , 2.
Most students were able to engage with the application independently and comfortably.
This familiarity allowed them to focus more on the speaking tasks rather than on learning how to use the tool itself.
If using ChatGPT for patent translation might change the rules for language use in Europe, and if so, what those rules will be (Larroyed, 2.
Nevertheless, some challenges were noted during the research process.
One issue was the variation in student engagement.
While many students used ChatGPT actively and meaningfully, a few tended to provide only short responses or relied too much on the AIAos input without reflecting on it.
This highlights the importance of researcher guidance to ensure that students use the tool effectively for practice rather than passively.
Another minor challenge was related to internet connectivity, which occasionally caused delays in response or interruptions during practice Although these interruptions were not widespread, they did affect the consistency of practice for some students.
Participants reported limited engagement mainly because they used ChatGPT for writing rather than for realPublished By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page time speaking tasks.
In contrast, most students in this study were actively involved in direct oral interaction with ChatGPT, though some still demonstrated dependency on the AIAos responses.
Thus, while all three studies observed obstacles in maximizing AI-assisted learning, the nature of the challenges varied depending on studentsAo digital readiness and the mode of ChatGPT integration during instruction (Setiawan et al.
, 2.
Overall, integrating ChatGPT proved to be a flexible and accessible tool for enhancing studentsAo speaking levels.
It offered immediate feedback, consistent practice opportunities, and a non-judgmental environment that encouraged students to speak more freely.
Compared to the control group, the experimental group displayed a higher level of spoken language proficiency, suggesting that AI tools like ChatGPT can be a valuable supplement to traditional speaking instruction, especially when students are already digitally literate and motivated to engage with technology.
The researcher supports previous studies that have shown the positive impact of integrating ChatGPT into the learning process on studentsAo speaking abilities.
This research successfully demonstrated that incorporating ChatGPT as a learning tool effectively enhanced studentsAo speaking abilities and increased their willingness to learn English at SMKN 1 Bulukumba.
CONCLUSION AND RECOMMENDATION
In conclusion, this research revealed a significant and positive enhancement in the speaking abilities of eleventh-grade vocational students majoring in accounting at SMKN 1 Bulukumba.
Enhancement was evident across all five assessed components of speaking ability.
Grammar achieved the highest improvement, followed by pronunciation, comprehension, fluency, and vocabulary.
These gains reflect balanced development, showing that ChatGPT-4oAos interactive features, real-time feedback, and voice-based conversation mode effectively addressed various aspects of speaking competence.
The integration created a dynamic and engaging learning environment where students could practice and refine their speaking skills beyond the limitations of traditional methods.
This approach enhanced grammatical accuracy, articulation, fluency, word choice, and comprehension, ultimately leading to greater confidence and communicative competence.
The findings highlight that ChatGPT provides a dynamic, interactive, and student-centered environment that promotes Through real-time conversation, and continuous engagement, students became more confident and fluent speakers.
The AIAos supportive tone and accessibility encouraged active participation, while the teacherAos presence ensured that learning remained guided and meaningful.
This balance between human instruction and AI support proved to be an effective model for improving EFL learnersAo oral proficiency.
For future researchers, it is recommended to extend the duration of the treatment to examine the long-term impact of ChatGPT on studentsAo speaking development.
Further studies could also explore its integration into other language skills such as listening, reading, and writing, or its application in different educational levels and learning environments.
Published By : CV.
Eureka Murakabi Abadi | https://jurnal-eureka.
com | Email :ijoleh.
journal@gmail.
Page REFERENCES