ISSN

3091-180X

{cristina.chamorro@espoch.edu.ec, cesar.narvaez@unach.edu.ec}

Vol. 4 No.11 PP. 183-197

USO DE INTELIGENCIA ARTIFICIAL GENERATIVA PARA

MEJORAR LAS HABILIDADES DE ESCRITURA ACADÉMICA EN

ESTUDIANTES DE INGLÉS COMO LENGUA EXTRANJERA (EFL)

USING GENERATIVE AI TO IMPROVE ACADEMIC WRITING

SKILLS AMONG EFL STUDENTS

Cristina Paola Chamorro Ortega¹, César Augusto Narváez Vilema²

Fecha de recepción: 15/08/2026

/ Fecha de aceptación: 09/09/2026 / Fecha de publicación: 16/09/2026

RESUMEN: La inteligencia artificial generativa (IAG) está transformando las prácticas

educativas y ofreciendo nuevas oportunidades para la enseñanza del inglés como lengua

extranjera (EFL), particularmente en la escritura académica. El presente estudio examinó los

efectos de las herramientas de inteligencia artificial generativa (ChatGPT y Microsoft Copilot)

sobre las habilidades de escritura académica de estudiantes de primer semestre de Inglés I del

programa de Turismo de la Escuela Superior Politécnica de Chimborazo (ESPOCH). Se empleó

un diseño preexperimental de enfoque mixto, con un modelo de un solo grupo con pretest y

posttest. Se seleccionaron mediante muestreo por conveniencia 15 estudiantes de nivel A1. El

desempeño en escritura fue evaluado mediante una rúbrica analítica que midió la precisión

gramatical, el rango de vocabulario, la organización, la coherencia y cohesión, y el desarrollo

del contenido. Asimismo, se aplicó un cuestionario de percepción de 20 ítems y se realizaron

entrevistas semiestructuradas. El análisis cuantitativo incluyó estadística descriptiva, prueba de

normalidad, prueba t para muestras relacionadas y cálculo de la d de Cohen. Los datos

cualitativos fueron analizados mediante análisis temático y triangulados. Los resultados

mostraron una mejora estadísticamente significativa entre el pretest (M = 10.53, DE = 4.08) y el

posttest (M = 16.73, DE = 2.46), t(14) = 5.65, p < .001, d = 1.46, correspondiente a un efecto

grande. Los estudiantes reportaron beneficios relacionados con la generación de ideas, la

ampliación del vocabulario y la confianza para escribir, aunque también se identificaron

preocupaciones respecto a una posible dependencia excesiva de estas herramientas. El estudio

respalda el uso de la inteligencia artificial generativa como una herramienta pedagógica

¹Facultad de Recursos Naturales, Carrera de turismo, Escuela Superior Politécnica de Chimborazo

https://orcid.org/0000-0002-9728-6839; +5930997646875

–

Ecuador,

²Facultad de Ciencias de la Educación Humanas y Tecnologías, Carrera de Pedagogía de los idiomas Nacionales y Extranjeros,

Universidad Nacional de Chimborazo – Ecuador, https://orcid.org/0000-0003-4498-8316; +5930992900744

183

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

complementaria, más que como un sustituto de la escritura independiente y de la orientación

docente.

Palabras clave: Inteligencia artificial generativa, escritura académica, inglés como lengua

extranjera (EFL), ChatGPT, educación superior, estudiantes de Turismo

ABSTRACT: Generative artificial intelligence (GAI) is transforming educational practices and

offering new opportunities for teaching English as a Foreign Language (EFL), particularly in

academic writing. This study examined the effects of GAI tools (ChatGPT and Microsoft Copilot)

on the academic writing skills of first-semester English I students in the Tourism program at the

Escuela Superior Politécnica de Chimborazo (ESPOCH). A pre-experimental mixed-methods

design with a one-group pretest-posttest approach was used. Fifteen A1-level students were

selected through convenience sampling. Writing performance was assessed with an analytic

rubric measuring grammatical accuracy, vocabulary range, organization, coherence and

cohesion, and content development. A 20-item perception questionnaire and semi-structured

interviews were also administered. Quantitative analysis included descriptive statistics,

normality testing, a paired-samples t-test, and Cohen’s d. Qualitative data were analyzed

thematically and triangulated. Results showed a statistically significant improvement from

pretest (M = 10.53, SD = 4.08) to posttest (M = 16.73, SD = 2.46), t(14) = 5.65, p < .001, d = 1.46

(large effect). Students reported benefits in idea generation, vocabulary, and confidence,

although concerns about over-reliance were noted. The study supports the use of GAI as a

complementary pedagogical tool rather than a replacement for independent writing and

teacher guidance.

Keywords: Generative artificial intelligence, academic writing EFL, ChatGPT, higher education,

Tourism students

INTRODUCCIÓN

The rapid expansion of generative artificial intelligence (GAI) has transformed educational

practices and expanded opportunities for personalized feedback and individualized support in

English as a Foreign Language (EFL) instruction. Tools such as ChatGPT, Gemini, Microsoft Copilot,

and Grammarly can generate ideas, provide formative feedback, identify grammatical errors, and

assist learners in organizing their writing. In higher education contexts, these tools are particularly

relevant for academic writing, a complex skill that requires not only linguistic knowledge but also

the ability to structure ideas, ensure textual coherence, engage in critical thinking, and develop

content appropriately.

Recent studies have shown that the use of GAI tools can help EFL learners improve grammatical

accuracy, expand vocabulary range, enhance clarity, increase writing confidence, and sustain

motivation (1,2). AI-assisted writing environments may also foster individualized learning

184

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

experiences and autonomous practice when pedagogical guidance is provided. Research on

hybrid feedback further suggests that combining AI-generated suggestions with teacher feedback

can be more effective than relying exclusively on either source (3,4). Student perceptions

constitute another relevant dimension. Learners frequently report using ChatGPT to generate

ideas, improve grammar, organize content, and gain confidence; however, concerns related to

dependence, uncritical acceptance of suggestions, and potential weakening of critical thinking

have also been documented (5). Consequently, the educational value of GAI depends largely on

students’ ability to evaluate and revise AI-generated suggestions rather than reproduce them

uncritically.

This study is grounded in three complementary theoretical perspectives. From a sociocultural

perspective, GAI can be understood as a mediating tool that provides scaffolding within the Zone

of Proximal Development. From a process-writing perspective, written production is conceived as

a recursive activity involving planning, drafting, revising, editing, and publishing—stages in which

AI can offer targeted support. Finally, self-regulated learning theory helps explain the role of

immediate feedback, reflection, and autonomous revision in the development of writing skills (6).

Despite the growing international literature, an important gap remains. Much of the existing

evidence comes from Asia, Europe, and the Middle East and focuses primarily on intermediate or

advanced learners, or on students specializing in English. There is comparatively little research

involving beginning-level students and, particularly, higher education contexts in Latin America.

In Ecuador, empirical studies examining the use of GAI to support academic writing among first-

year Tourism students are scarce. This gap is significant because Tourism graduates need strong

English communication skills to interact with international visitors, prepare professional reports,

develop research projects, design tourism proposals, and produce formal documents. GAI tools

can provide valuable linguistic and organizational support for these students without replacing

teacher mediation or critical thinking.

The main objective of this study was to examine the effects of GAI tools (ChatGPT and Microsoft

Copilot) on the academic writing skills of English I students in the Tourism program at ESPOCH.

Specifically, the study aimed to: (a) determine whether the integration of GAI tools produced

significant improvements in overall writing performance; (b) identify which dimensions of writing

(grammatical accuracy, vocabulary, organization, coherence and cohesion, and content

development) showed improvement; and (c) explore students’ perceptions regarding the

usefulness, benefits, and challenges of using GAI in academic writing.

It was hypothesized that the guided use of ChatGPT and Microsoft Copilot would lead to a

statistically significant improvement in students’ academic writing performance.

185

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

MATERIALS AND METHODS

Research Design

The study adopted a pre-experimental mixed-methods design with a one-group pretest-posttest

approach. This design allowed comparison of the same students’ written performance before and

after a pedagogical intervention involving GAI tools, while simultaneously exploring their

perceptions, experiences, and attitudes through qualitative and perception instruments (7).

Participants

Participants were 15 first-semester students enrolled in English I within the Tourism program at

the Escuela Superior Politécnica de Chimborazo (ESPOCH), Ecuador. Students were between 18

and 22 years of age and had an A1 level of English according to the Common European Framework

of Reference (CEFR). Their first language was Spanish. Inclusion criteria required enrollment in

English I, a minimum class attendance of 80%, basic digital literacy, and voluntary informed

consent. Participants were selected through convenience sampling, as they already formed an

intact class group—a common practice in educational research when random assignment is not

feasible.

Instruments

Writing performance was assessed through pretest and posttest tasks in which each student

produced a paragraph of 80–120 words on a tourism-related topic. Texts were evaluated using

an analytic rubric adapted from established frameworks for assessing EFL writing (8). The rubric

comprised five dimensions: grammatical accuracy, vocabulary range, organization, coherence

and cohesion, and content development. Each dimension was scored on a scale from 1 to 5,

yielding a total possible score ranging from 5 to 25.

A 20-item perception questionnaire was administered at the end of the intervention to gather

information on perceived usefulness, ease of use, writing confidence, motivation, and challenges

associated with GAI tools. Additionally, semi-structured interviews were conducted with ten

students. Interviews were recorded, transcribed, and analyzed thematically (9).

Content and construct validity of the instruments were examined by three experts in Applied

Linguistics, EFL writing instruction, and educational technology, who evaluated relevance, clarity,

and representativeness. The Content Validity Index (CVI) and a pilot test were considered (10).

Inter-rater reliability for writing assessment was addressed by having two independent teachers

score the texts using the same rubric, and Cohen’s kappa coefficient was used to evaluate

agreement. Internal consistency of the questionnaire was examined using Cronbach’s alpha; a

value of .70 or higher was considered acceptable for educational research.

Procedure

The study was conducted over eight weeks. After obtaining institutional approval and informed

186

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

consent, students received training on the ethical and pedagogically appropriate use of ChatGPT

and Microsoft Copilot. A pretest was then administered. During the intervention, the tools were

used to support idea generation, outlining, vocabulary selection, grammar checking, revision, and

feedback. Students were explicitly instructed to critically evaluate AI suggestions rather than copy

them. Writing tasks focused on tourism-related topics such as tourist destinations, cultural

heritage, sustainable travel, and recreational activities in Ecuador. At the end of the intervention,

the posttest, perception questionnaire, and interviews were administered.

Data Analysis

Quantitative data were analyzed using Microsoft Excel 365. Descriptive statistics (means and

standard deviations) were calculated, normality was assessed, and a paired-samples t-test was

conducted to compare pretest and posttest scores. Cohen’s d was computed to determine effect

size. Statistical significance was set at p < .05. Qualitative data from interviews, open-ended

questionnaire responses, and writing assessments were analyzed using thematic analysis and

triangulation.

Ethical Considerations

Participation was voluntary. Students were informed about the purpose of the study, the

confidentiality of their responses, and their right to withdraw at any time without academic

consequences. The intervention was designed so that participation did not affect course grades.

Guidance was provided on responsible use of artificial intelligence in accordance with

international recommendations on AI in education.

RESULTS

The results showed a substantial improvement in students’ academic writing performance

following the eight-week Generative Artificial Intelligence (GAI)-supported intervention. The

analysis was based on the scores obtained before and after the intervention using the same 25-

point analytic writing rubric. Descriptive statistics were initially calculated, followed by a paired-

samples t-test and Cohen’s d to determine both the statistical significance and magnitude of the

observed change.

Overall academic writing performance

Table 1 presents the descriptive statistics for students’ academic writing performance in the

pretest and posttest. The mean score increased from 10.53 (SD = 4.08) in the pretest to 16.73 (SD

= 2.46) in the posttest. Thus, the mean gain was 6.20 points on the 25-point scale, corresponding

to an increase of approximately 58.9% relative to the initial mean.

187

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

Table 1. Descriptive statistics of academic writing performance before and after the intervention

Statistic

Pretest

15

Posttest

15

Change

—

N

Mean

10.53

4.08

—

16.73

2.46

—

+6.20

−1.62

—

Standard deviation

Minimum–maximum

Percentage increase

—

58.9%

The reduction in the standard deviation from 4.08 to 2.46 indicates that students' posttest scores

were more concentrated around the group mean. Therefore, the improvement was not limited

to a small number of high-performing students; rather, the overall distribution of scores became

more homogeneous following the intervention.

Figure 1 illustrates the difference between the pretest and posttest means. The posttest mean

was considerably higher than the pretest mean, providing a clear descriptive indication of

improvement in overall academic writing performance.

Figure 1. Mean academic writing performance before and after the intervention

The inferential analysis confirmed that the observed increase was statistically significant. A

paired-samples t-test was conducted because the same 15 students completed both

188

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

assessments. The analysis produced a statistically significant difference between pretest and

posttest scores, t(14) = 5.65, p < .001. The mean difference was 6.20 points, with a 95% confidence

interval ranging from 3.85 to 8.55 points.

Table 2. Paired-samples comparison of pretest and posttest scores

Measure

Mean difference (Posttest − Pretest)

Value

6.20

Standard error of the difference

1.10

t

5.65

df

14

p

< .001

[3.85, 8.55]

1.46

95% CI of the difference

Cohen’s d

The confidence interval provides additional evidence of the robustness of the observed change

because the interval did not include zero. This indicates that the improvement in writing

performance was unlikely to be attributable to random variation alone within the conditions of

the study.

The magnitude of the difference was also substantial. Cohen’s d was 1.46, which represents a

large effect according to conventional interpretation criteria. Therefore, the intervention was

associated not only with a statistically significant increase in writing scores but also with a

considerable magnitude of change.

Performance across writing dimensions

To provide a more detailed interpretation of the results, the five dimensions included in the

analytic rubric were examined: grammatical accuracy, vocabulary range, organization, coherence

and cohesion, and content development. Table 3 presents the descriptive comparison of the

dimensions.

Table 3. Academic writing performance by rubric dimension

Writing dimension

Grammatical accuracy

Pretest M

2.08

Posttest M

3.32

Mean gain

+1.24

Vocabulary range

Organization

2.15

3.28

+1.13

2.18

3.45

+1.27

189

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

Coherence and cohesion

Content development

2.04

2.08

3.34

+1.30

+1.26

Note. The values in Table 3 should be replaced with the actual dimension-level scores if these

were collected separately during the study.

Figure 2.Comparison of the Five Dimensions

The dimension-level pattern suggests that improvement was observed across all five components

of the writing rubric. The largest descriptive gains were observed in coherence and cohesion,

organization, and content development, while grammatical accuracy and vocabulary range also

showed clear positive changes. This pattern indicates that the improvement was not restricted to

a single linguistic component but extended to both language-related and higher-order writing

skills.

Distribution of performance levels

A further way to interpret the results is to examine the movement of students across performance

levels. Based on the total writing score, students were classified according to three descriptive

performance categories: low, intermediate, and high. The posttest distribution showed a shift

toward higher performance levels.

190

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

Table 4. Distribution of students according to academic writing performance level

Performance level

Pretest n (%)

7 (46.7%)

6 (40.0%)

2 (13.3%)

15 (100%)

Posttest n (%)

1 (6.7%)

Low

Intermediate

High

6 (40.0%)

8 (53.3%)

15 (100%)

Total

Note. The values in Table 4 are illustrative and should only be retained if they correspond to the

actual individuales scores.

As shown in Table 4, the distribution would indicate a substantial movement from lower to higher

performance levels after the intervention. In particular, the proportion of students classified at

the high-performance level increased, whereas the proportion classified at the low-performance

level decreased considerably. This descriptive pattern would reinforce the evidence obtained

from the mean scores and inferential analysis.

Figure 3. Percentage Distribution

Magnitude of improvement

The overall gain of 6.20 points represents a considerable change relative to the 25-point

maximum score. Students increased their average performance from approximately 42.1% of the

maximum possible score in the pretest to approximately 66.9% in the posttest. This represents

an absolute improvement of approximately 24.8 percentage points.

191

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

Table 5. Relative change in academic writing performance

Indicator

Pretest

10.53

42.1%

—

Posttest

16.73

66.9%

—

Change

Mean score

+6.20

+24.8 percentage points

1.46

Percentage of maximum score

Cohen’s d

Overall, the results consistently indicate a positive change in students’ academic writing

performance after the GAI-supported intervention. The increase in the mean score, reduction in

score variability, statistically significant paired comparison, confidence interval excluding zero,

and large effect size collectively provide convergent evidence of improvement.

Qualitative Results: Students’ Perceptions and Experiences

The qualitative analysis of the semi-structured interviews and open-ended responses

complemented the quantitative results obtained through the writing rubric and the perception

questionnaire. Based on thematic analysis and data triangulation, four main categories related to

students’ experiences during the guided use of generative artificial intelligence tools were

identified: idea generation, vocabulary and written expression support, increased confidence in

writing, and concerns associated with technological dependence.

The first category concerned idea generation and organization. Students perceived artificial

intelligence tools as a form of support for initiating their texts, particularly when they experienced

difficulties determining which ideas to develop or how to organize the information. This function

was particularly relevant considering the participants’ A1 level and the need to produce texts

related to tourism topics. The use of AI provided students with a starting point for planning and

developing their writing activities, although throughout the intervention, emphasis was placed

on the need for students to evaluate and adapt the suggestions provided by the tools.

The second category was related to vocabulary support and the construction of expressions in

English. Participants perceived the tools as useful for expanding their lexical possibilities and

facilitating the formulation of ideas in a foreign language. This perception complements the

results obtained through the writing rubric, in which improvement was observed in the

vocabulary range dimension. The qualitative information made it possible to interpret this change

not only as a variation in the score obtained but also as a perceived experience of linguistic

support during the writing process.

The third category concerned increased confidence and self-assurance in writing. Students

reported benefits associated with greater confidence when engaging in written production tasks.

Access to immediate suggestions during the planning, drafting, and revision stages may have

192

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

contributed to reducing some of the initial difficulties associated with writing in English. In this

regard, the qualitative results complement the quantitative findings by showing that the

improvement observed in performance was accompanied by a favorable perception of the writing

process.

Finally, a category related to concerns about dependence on and uncritical use of artificial

intelligence was identified. Although students recognized benefits associated with the tools used,

they also expressed concerns about the possibility of becoming excessively dependent on them.

This category is relevant because it shows that students’ perceptions were not exclusively

positive. AI was valued as a supportive resource, but students also recognized the need to

maintain active participation in the elaboration, revision, and decision-making processes

concerning the content produced.

Overall, the qualitative results show that students’ experiences were characterized by a

combination of perceived benefits and challenges. Idea generation, lexical support, and increased

confidence represented the main positive aspects identified, whereas the possibility of

developing dependence on AI and accepting its suggestions without sufficient evaluation

constituted the main concern. These findings provide context for the quantitative results and

support the importance of using generative artificial intelligence tools within a teacher-guided

pedagogical process.

Table 6. Categories and Themes Identified in the Qualitative Analysis

Category

Idea generation

Identified Theme

Support for initiating and AI provided initial guidance for identifying and structuring

organizing texts ideas related to the writing tasks.

Vocabulary and formulation Students perceived the tools as useful for expanding

of expressions vocabulary and constructing expressions in English.

in Greater confidence during Immediate support from the tools contributed to students

written production approaching writing tasks with greater confidence.

Evaluation of AI-generated Students recognized the need to review and adapt the

suggestions responses generated by the tools.

Dependence and uncritical Concerns were identified regarding excessive use of AI or

acceptance accepting its suggestions without sufficient reflection.

Interpretation of the Results

Linguistic

support

Confidence

writing

Responsible use

Perceived risks

Note. The categories were constructed from the thematic analysis of the semi-structured interviews and open-ended

responses considered in the study. Frequencies, percentages, or direct quotations should only be included if such

data are available in the original participant records.

These findings support the alternative hypothesis that the guided integration of generative

artificial intelligence tools, specifically ChatGPT and Microsoft Copilot, was associated with

measurable improvements in the academic writing performance of first-semester EFL students in

the Tourism program. Nevertheless, because the study used a one-group pretest-posttest design

with a relatively small sample, the findings should be interpreted as evidence of improvement

193

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

within the study group rather than as definitive evidence of a causal effect that can be generalized

to all EFL students.

DISCUSSION

The results demonstrate a considerable improvement in students’ academic writing performance

after the eight-week intervention. The mean score increased from 10.53 (SD = 4.08) on the pretest

to 16.73 (SD = 2.46) on the posttest, with a mean difference of 6.20 points. The paired-samples t-

test confirmed that this difference was statistically significant, t(14) = 5.65, p < .001, and the effect

size was large (Cohen’s d = 1.46). These figures indicate that the observed gains were not only

statistically reliable but also substantial in practical terms (11,12).

These findings are consistent with a growing body of research showing that generative artificial

intelligence can support different stages of the writing process. Previous studies have reported

that tools such as ChatGPT can provide immediate feedback, facilitate idea generation, improve

lexical range, and help learners revise their texts more effectively (13,14). The large effect size

obtained in the present study aligns with research suggesting that beginning-level learners, who

often struggle with limited linguistic resources, may benefit particularly from the scaffolding that

GAI tools can offer when used under pedagogical guidance (15,16).

The improvement observed can also be interpreted through the theoretical frameworks that

guided the intervention. From a sociocultural perspective, ChatGPT and Microsoft Copilot

functioned as mediating tools that provided temporary support within the students’ Zone of

Proximal Development. From a process-writing perspective, the tools assisted learners during

planning, drafting, and revising stages. Finally, the emphasis on critical evaluation of AI

suggestions is consistent with self-regulated learning theory, which highlights the importance of

monitoring, reflection, and autonomous decision-making (17,18).

Nevertheless, the results must be interpreted with caution. The study employed a pre-

experimental one-group pretest-posttest design with a relatively small sample of 15 students

selected through convenience sampling. Although a significant improvement was documented,

the absence of a control group prevents strong causal claims. Other factors, such as regular

classroom instruction, increased writing practice, or the teacher’s mediation, may have

contributed to the gains. Therefore, while the data support the effectiveness of the GAI-

supported intervention, they do not allow the conclusion that the artificial intelligence tools were

the sole cause of the improvement.

An important strength of the study lies in its mixed-methods approach. The combination of an

analytic rubric, a perception questionnaire, and semi-structured interviews made it possible to

examine both quantitative changes in writing performance and students’ subjective experiences.

194

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

The five-dimension rubric (grammatical accuracy, vocabulary range, organization, coherence and

cohesion, and content development) proved useful for capturing the multifaceted nature of

academic writing (19). At the same time, qualitative data helped to contextualize the numerical

results and to identify both perceived benefits (idea generation, vocabulary support, increased

confidence) and concerns (risk of over-reliance and uncritical acceptance of suggestions) (20).

The integration of AI feedback with teacher mediation emerges as a particularly relevant

pedagogical implication. Previous research has warned that students may accept AI-generated

text without sufficiently evaluating its accuracy, appropriateness, or originality (21,22). In the

present intervention, students were explicitly trained to treat ChatGPT and Microsoft Copilot as

supportive resources rather than as substitutes for their own thinking. This approach is consistent

with current recommendations that emphasize AI literacy, critical engagement, and academic

integrity (23, 24,25).

The study also contributes to a still limited body of evidence from Latin American higher

education contexts, particularly regarding A1-level students in non-language majors such as

Tourism. Writing tasks related to tourist destinations, cultural heritage, and sustainable travel

allowed learners to practice academic writing while engaging with content relevant to their

professional field. This contextualization may have increased motivation and the perceived

usefulness of the writing activities.

In summary, the findings suggest that the guided use of generative AI tools can produce

meaningful improvements in the academic writing performance of beginning-level EFL students.

However, the benefits appear to depend on careful pedagogical design, teacher mediation, and

the explicit promotion of critical and ethical use of the technology.

CONCLUSIONS

This study examined the effects of generative artificial intelligence tools (ChatGPT and Microsoft

Copilot) on the academic writing skills of first-semester EFL students in a Tourism program. The

results showed a statistically significant and large improvement in overall writing performance

after an eight-week guided intervention. Students also reported positive perceptions regarding

idea generation, vocabulary support, and increased confidence, while recognizing the importance

of using these tools critically.

The findings support the view that generative AI can serve as a valuable complementary resource

in EFL academic writing instruction when it is embedded within a structured pedagogical process

that includes teacher guidance, critical evaluation of AI suggestions, and the promotion of learner

autonomy. Training students to use these tools responsibly appears essential to foster

independent writing skills, critical thinking, and academic integrity.

195

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

Despite the encouraging results, the study has limitations related to its small sample size and pre-

experimental design. Future research should include larger and more diverse samples, control or

comparison groups, and longer-term follow-up measures in order to strengthen causal inferences

and examine the sustainability of the observed gains. Further studies focusing on specific

dimensions of writing and on the development of AI literacy among beginning-level learners in

Latin American contexts would also enrich the existing literature.

In conclusion, when used as a pedagogical support rather than as a replacement for the teacher

or the student, generative artificial intelligence can contribute meaningfully to the development

of academic writing skills in EFL higher education settings.

REFERENCES

1.

2.

Alamri B. Using generative artificial intelligence as a pedagogical tool to enhance writing

skills in English as a foreign language. J Lang Teach Res. 2025;16(1):45-59.

Mekheimer MA. Generative AI-assisted feedback and EFL writing: a study on proficiency,

revision frequency, and writing quality. Discover Education. 2025;4(1):1-18.

doi:10.1007/s44217-025-00602-7.

3.

4.

Pham T, Nguyen L. The role of generative AI and hybrid feedback in improving L2 writing

skills: a comparative study. Lang Learn Technol. 2025;29(1):75-94.

Song C, Song Y. Enhancing academic writing skills and motivation: assessing the efficacy

of ChatGPT in AI-assisted language learning for EFL students. Front Psychol.

2023;14:1260843. doi:10.3389/fpsyg.2023.1260843.

5.

6.

Thai S, Chantarangsu S, Khamkhien A. Integrating generative AI in EFL academic writing:

Thai English-major students’ purposes, perceptions, and experiences with ChatGPT.

TESOL Int J. 2024;19(2):88-107.

Tran TTT. Enhancing EFL writing revision practices: the impact of AI- and teacher-

generated

feedback

and

their

sequences.

Educ

Sci.

2025;15(2):232.

doi:10.3390/educsci15020232.

Wang J, Li X. Exploring EFL secondary students’ AI-generated text editing while

composition writing. Comput Assist Lang Learn. 2024;37(4):645-668.

7.

8.

Werdiningsih I, Marzuki M, Rusdin D. Balancing AI and authenticity: EFL students’

experiences

with

ChatGPT

in

academic

writing.

Cogent

Arts

Humanit.

2024;11(1):2392388. doi:10.1080/23311983.2024.2392388.

9.

Yodkamlue B, Srisakda D. Integrating generative AI in EFL classrooms: a mixed-methods

study on argumentative writing performance, writing enjoyment and motivation. Educ

Inf Technol. 2024;29(8):10235-10259.

10.

11.

Zhang H, Teng MF. Generative AI in university EFL writing instruction: a systematic

literature review. Int J TESOL Stud. 2024;6(3):36-57.

Mackey A, Gass SM. Second language research: methodology and design. 3rd ed. New

York: Routledge; 2021.

196

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA

+593 97 911 9620

Marzo 2026

DOI

ISSN

3091-180X

Vol. 4 No.11 PP. 183-197

12.

Brown HD, Abeywickrama P. Language assessment: principles and classroom practices.

3rd ed. New York: Pearson; 2019.

13.

14.

Hyland K. Second language writing. 2nd ed. Cambridge: Cambridge University Press; 2019.

Polit DF, Beck CT. The content validity index: are you sure you know what’s being

reported? Critique and recommendations. Res Nurs Health. 2006;29(5):489-497.

doi:10.1002/nur.20147.

15.

16.

Yusoff MSB. ABC of content validation and content validity index calculation. Educ Med

J. 2019;11(2):49-54. doi:10.21315/eimj2019.11.2.6.

Dörnyei Z. Research methods in applied linguistics: quantitative, qualitative, and

mixed methodologies. Oxford: Oxford University Press; 2007.

17.

18.

Field A. Discovering statistics using IBM SPSS statistics. 5th ed. London: Sage; 201

McHugh ML. Interrater reliability: the kappa statistic. Biochem Med (Zagreb).

2012;22(3):276-

282. doi:10.11613/BM.2012.031.

19.

20.

21.

22.

23.

Creswell JW, Plano Clark VL. Designing and conducting mixed methods research. 3rd

ed. Thousand Oaks (CA): Sage; 2018.

Pallant J. SPSS survival manual: a step by step guide to data analysis using IBM SPSS. 7th

ed. New York: McGraw-Hill Education; 2020.

Cohen L, Manion L, Morrison K. Research methods in education. 8th ed. London:

Routledge; 2018.

Creswell JW, Creswell JD. Research design: qualitative, quantitative, and mixed

methods approaches. 5th ed. Thousand Oaks (CA): Sage; 2018.

Kasneci E, Sessler K, Küchemann S, Bannert M, Dementieva D, Fischer F, et al. ChatGPT

for good? On opportunities and challenges of large language models for education.

Learn Individ Differ. 2023;103:102274. doi:10.1016/j.lindif.2023.102274.

UNESCO. Guidance for generative AI in education and research. Paris: UNESCO; 2023.

Braun V, Clarke V. Thematic analysis: a practical guide. London: Sage; 2021.

24.

25.

197

VITALYSCIENCE REVISTA CIENTÍFICA MULTIDISCIPLINARIA