DOI: https://doi.org/10.71112/xy8ca073
708 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
The use of Gemini IA to enhance the english language speaking skill
El uso de Gemini AI para mejorar la destreza del habla inglesa
Miguel Angel Miguez Gordillo
a,*
soiyomiguel@gmail.com
https://orcid.org/0009-0002-2217-3706
*
Autor de correspondencia: soiyomiguel@gmail.com,
a
ISU SUCRE, Ecuador
ABSTRACT
The integration of Artificial Intelligence (AI) into English as a Foreign Language (EFL) teaching
represents a pedagogical paradigm shift. This research analyzes the influence of Gemini-AI on
the development of speaking skills in 15 A2-level students at the Instituto Universitario Sucre.
Using a qualitative approach with a descriptive scope, observation instruments were employed
to assess student perceptions and the tool's technical impact. The theoretical framework is
based on language acquisition through comprehensible input and simulated immersion
environments. The results indicate that, after 30 days of intermittent use, participants increased
their speaking performance by 10% compared to their baseline. The AI was identified as an
effective scaffolding resource, immediately detecting gaps in vocabulary, grammar, and
collocations. The main conclusion highlights that the non-punitive nature of the interaction with
the AI functions as a "communicative warm-up" that significantly reduces language anxiety. The
adoption of Gemini-AI is recommended as a robust and ethical educational resource capable of
fostering effective knowledge transfer in contemporary educational settings.
Keywords: Artificial Intelligence; language teaching; speech; language anxiety; Gemini-AI;
pedagogical scaffolding
DOI: https://doi.org/10.71112/xy8ca073
709 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
RESUMEN
La integración de la Inteligencia Artificial (IA) en la enseñanza del inglés como lengua
extranjera (EFL) representa un cambio de paradigma pedagógico. Esta investigación analiza la
influencia de Gemini-AI en el desarrollo de la producción oral (speaking) en 15 estudiantes de
nivel A2 del Instituto Universitario Sucre. Bajo un enfoque cualitativo con alcance descriptivo, se
emplearon instrumentos de observación para evaluar la percepción estudiantil y el impacto
técnico de la herramienta. El marco teórico se fundamenta en la adquisición de lenguas
mediante inputs comprensibles y entornos de inmersión simulada. Los resultados indican que,
tras 30 días de uso intermitente, los participantes incrementaron su desempeño oral en un 10%
respecto a la línea base. Se identificó que la IA actúa como un recurso de andamiaje
(scaffolding) eficaz, detectando brechas en vocabulario, gramática y colocaciones de forma
inmediata. La conclusión principal destaca que la naturaleza no punitiva de la interacción con la
IA funciona como un "calentamiento comunicativo" que reduce significativamente la ansiedad
lingüística. Se recomienda la adopción de Gemini-AI como un recurso didáctico sólido y ético,
capaz de propiciar una transferencia de conocimiento efectiva en escenarios educativos
contemporáneos.
Palabras clave: Inteligencia Artificial; enseñanza de lenguas; speaking; ansiedad lingüística;
Gemini-AI; andamiaje pedagógico
Received: April 1, 2026 | Accepted: April 19, 2026 | published: April 20, 2026
DOI: https://doi.org/10.71112/xy8ca073
710 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
INTRODUCTION
Education has changed since its origin because it is its nature to be a changing process,
it means to adapt, modify, evolve, according to the necessities of students and society. It makes
gradual changes that appear to update methodologies, material, modality, instruments and
processes, but also the human actors like teachers, students, and academic community evolve
at the same time. Taking this reality into account, we have to implement new methodologies and
techniques that adapt to the actual society (Benyo, 2020).
Nowadays, we live in a society that is more virtual and technological than before. As a
result of this reality, we have to employ technology as one of our main academic tools, with the
objective to be one of our main helpers to continue enhancing the education process. In this
sense, Artificial Intelligence (AI) has emerged as a main actor in multiple fields, and education is
no exception. AI refers to the development of computer systems that can perform tasks which
require human intelligence, such as problem-solving, language understanding, and decision-
making (Hernando Barrios-Tao, 2026). The evolution of IA has made meaningful changes in
society. In the education environment, AI is revolutionizing traditional learning rooms by offering
personalized learning experiences, automating administrative tasks, and providing intelligent
tutoring systems. As classrooms become more digital and connected, the integration of AI
presents both exciting opportunities and important challenges. AI has the power to transform the
learning process because it can adapt to student necessities and offer them a personalized
program or personalized guide to study, reinforcing the parts of a lesson that they most need
and avoiding what they have already known, improving the monitoring of student performance
(Alexandara Harry, 2023).
The integration of AI in language education has shifted from closed systems (CALLs) to
open and generative models. Following Creswell, this study adopts a theoretical perspective
where technology is not just a tool, but a socio-cognitive mediator (Benyo, 2020). As a result,
that study sustains that Google Gemini represents the pinnacle of this evolution. Its multimodal
DOI: https://doi.org/10.71112/xy8ca073
711 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
nature allows students not only to receive written input, but also to interact in a feedback loop of
"linguistic safe environment". Booth (zammit 2009) points out in “The Craft of Research”, the
validity of a scientific argument depends on evidence; in this context, Gemini provides dynamic
data, information, and knowledge that adapts to the student's Zone of Proximal Development
(ZPD), enabling speaking to evolve from rote repetition to creative production (Abbas Ali
Rezaee, 2012).
English communicative competence is a cornerstone for EFL learners, as it facilitates
the oral expression of complex thoughts and ideas through the precise integration of
pronunciation, grammar, and lexical naturalness. However, a critical paradox exists at the Sucre
Institute: despite years of formal instruction in primary and secondary education, students
across various disciplines continue to exhibit significant deficiencies in fluency and confidence.
This lack of proficiency is often exacerbated by limited opportunities for authentic immersion
within the local context and the socioeconomic barriers to international travel, leading to a
perceived lack of interest in the mandatory A1 and A2 levels. Notwithstanding these traditional
hurdles, this study addresses this pedagogical gap by investigating the impact of Gemini-AI
input on the development of listening and speaking skills. Specifically, the research evaluates:
(1) the efficacy of Gemini-generated texts in promoting the noticing of vocabulary, collocations,
and formulaic phrases; and (2) the correlation between this initial linguistic noticing and the
subsequent advancement of daily communicative fluency among foreign language students.
This study focuses on the strong demand and need to transform language teaching in
technical institutes through the use of disruptive tools. Theoretically, it explores how the use of
Gemini-AI fosters the conscious perception of grammatical and lexical structures in beginner
students (Miftakul Rokhman Purnama, 2025). In practice, the research introduces an accessible
methodology that compensates for the lack of contact with native speakers in the local
environment, unlocking academic progress at the basic level of the Instituto Sucre. The
DOI: https://doi.org/10.71112/xy8ca073
712 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
methodological value lies in the validation of prompts designed to optimize speaking and
listening skills, linking cutting-edge technology with the goals of the international curriculum.
This work is based on the multiple studies based on the AI implementation and the
strong demand to use new technological tools such as the "Generative Artificial Intelligence
tools” to provide simulated immersion environments that reduce linguistic anxiety and offer
highly personalized input. To comprehend this phenomenal, this research has been organized
in different sections with the purpose to reach the research objectives of Gemini use in
language learning and its future applications.
DEVELOPMENT
ARTIFICIAL INTELLIGENCE IN LANGUAGE EDUCATION: FROM CALL TO GENERATIVE
AI
The integration of technology in language learning is not a new phenomenon. For
decades, the field of Computer-Assisted Language Learning (CALL) has evolved in response to
new technological tools and the changing needs of learners. Initially, CALL systems were
primarily behaviorist, focused on repetitive grammar drills, vocabulary memorization, and
automated correction. These early systems were limited in interaction and failed to represent
the communicative nature of real language use. As digital environments progressed, CALL
shifted toward communicative approaches, supporting interaction, meaningful tasks, and
learner-centered environments (Benyo, 2020). However, even advanced digital learning tools
had a fundamental limitation: they could not fully simulate authentic conversation, especially in
contexts where learners had minimal exposure to native speakers or real immersion
opportunities. This limitation is especially evident in EFL environments such as technical
institutes, where students often have restricted contact with English beyond the classroom
(Guendouz, 2025).
The emergence of Generative Artificial Intelligence, particularly Large Language Models
(LLMs) such as Google Gemini, represents a new phase in the evolution of CALL. Unlike rule-
DOI: https://doi.org/10.71112/xy8ca073
713 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
based systems, LLMs generate language dynamically and respond to learners with near-human
fluency (Omar Mahmoud ELSenbawy, 2025). This is pedagogically significant because
speaking is not simply a matter of recalling grammar rules; speaking requires real-time
processing, vocabulary selection, coherence, confidence, and interactional competence
(Guendouz, 2025).
In this sense, Gemini is not merely a digital tool that “provides information.” Instead, it
functions as a language environment: a space where learners can engage in interaction,
produce output, receive feedback, and repeat communicative attempts without the social
pressure of human evaluation (Guendouz, 2025). This is particularly relevant for beginner-level
learners (A1–A2), whose communicative anxiety and limited linguistic resources often prevent
them from practicing speaking consistently. Therefore, the theoretical foundation of this
research is anchored in the view that Generative AI is not simply an external aid. It is a
mediating technology that reshapes the conditions of learning by providing continuous input,
scaffolding output, and supporting the learner’s emotional readiness for communication (Miftakul
Rokhman Purnama, 2025).
SOCIO-CONSTRUCTIVISM AND AI AS A MEDIATOR OF LEARNING
A key theoretical lens for understanding AI in education is socio constructivism, which
argues that learning occurs through interaction, social engagement, and meaning-making rather
than passive reception of information. From this perspective, knowledge is constructed through
dialogue, feedback, and negotiation of meaning. In language learning, this becomes even more
relevant because language itself is inherently social (Katikela Kishore, 2024).
Traditional language classrooms, particularly in EFL contexts, face structural limitations:
• Large groups limit individual speaking time
• Teachers cannot provide constant individualized feedback
• Students fear judgment, especially at beginner levels
• Interaction opportunities are scarce outside class
DOI: https://doi.org/10.71112/xy8ca073
714 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
Socio-constructivist theory emphasizes that learners progress when they engage in
guided interaction with more capable partners. This is strongly connected to Vygotsky’s Zone of
Proximal Development (ZPD). The ZPD refers to the gap between what a learner can do
independently and what they can do with assistance. In language learning, this assistance often
takes the form of scaffolding: correction, modeling, paraphrasing, vocabulary suggestions, and
strategic support (Abbas Ali Rezaee, 2012).
In traditional classrooms, teachers provide scaffolding, but their time is limited. Gemini
can function as a simulated “more capable peer”, offering real-time support that helps learners
operate within their ZPD. For example, when a student attempts to describe a routine, Gemini
can:
• reformulate incorrect sentences
• suggest more natural collocations
• provide vocabulary alternatives
• ask follow-up questions to maintain interaction
This interaction creates a dynamic feedback loop. The learner produces language,
receives scaffolding, modifies output, and gradually internalizes more accurate and fluent forms.
In socio-constructivist terms, Gemini becomes a mediator of learning, enabling interaction that
would otherwise be inaccessible in resource-limited environments (Hasanein, Sobaih, &
Elshaer, 2024). However, it is essential to clarify that Gemini does not replace the teacher’s role
as mediator. Instead, it extends mediation beyond classroom time. The teacher remains
responsible for designing tasks, ensuring alignment with learning outcomes, and guiding
students toward critical and ethical use. Thus, socio-constructivism supports the central
argument of this study: that Gemini can provide structured, interactive, and scaffolded practice,
which promotes the development of speaking skills through guided meaning making (Miftakul
Rokhman Purnama, 2025).
DOI: https://doi.org/10.71112/xy8ca073
715 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
INPUT, NOTICING, AND THE DEVELOPMENT OF COMMUNICATIVE COMPETENCE
One of the most influential bodies of theory in Second Language Acquisition (SLA)
concerns the role of input and the mechanisms through which learners transform input into
acquisition. A foundational concept is comprehensible input, which refers to language that
learners can understand while being slightly above their current level. Yet, research has
consistently shown that input alone is not enough (Krashen, 2020).
Learners must also develop the ability to consciously detect language features. This is
explained by the Noticing Hypothesis, which states that learners must notice vocabulary,
grammar patterns, and functional expressions in order to internalize them (Fitria, 2025). In
beginner-level EFL contexts, students often fail to notice key structures because:
• classroom exposure is limited
• instruction is often textbook-centered
• learners lack repeated interaction with authentic language
• feedback is delayed or insufficient
Gemini strengthens the noticing process in several ways:
a) Salience and repetition
Gemini can highlight vocabulary, connectors, collocations, and formulas repeatedly in
conversation. This repetition increases salience and helps learners recognize patterns (Eko S,
2025).
b) Contextualized learning
Unlike traditional exercises, Gemini provides language in context. For example, phrases such
as: “As far as I’m concerned”, “In my opinion”, “The main reason is” appear naturally within a
communicative situation. This strengthens functional competence.
DOI: https://doi.org/10.71112/xy8ca073
716 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
LEARNER CONTROL
Students can ask for explanations, synonyms, simpler versions, or more advanced
versions. This transforms passive input into active engagement. This noticing process is critical
for developing communicative competence (M. Rashtchi, 2018).
According to the CEFR framework, speaking competence includes: fluency and
coherence, lexical range, grammatical accuracy, interactional strategies. Gemini contributes to
all these areas by providing a consistent and adaptive conversational environment. It supports
learners in moving from isolated grammar knowledge toward real communication. In short,
Gemini serves as a tool that enhances noticing, which supports the transformation of input into
intake, and eventually into communicative output (Miew Luan Ng, 2025).
OUTPUT, INTERACTION, AND FLUENCY DEVELOPMENT
Another key principle in SLA is that language learning requires not only input but also
output. The Output Hypothesis argues that learners develop competence when they are forced
to produce language and realize gaps in their knowledge. Speaking is not simply a
demonstration of what learners know; speaking is a mechanism through which they learn
(Swain, 2005).
Gemini increases output opportunities dramatically. Instead of waiting for classroom
speaking turns, learners can practice: daily routines, roleplays, interviews, debates, problem-
solving dialogues (Purnama, 2025). This matters because fluency develops through repetition,
automation, and reduced cognitive effort. Fluency requires learners to access vocabulary and
structures quickly, without excessive mental struggle (Katikela Kishore, 2024). This process
connects directly with Cognitive Load Theory, which explains how learning is affected by the
limitations of working memory (Siti Nurul Azizah, 2025). When learners speak, they must
simultaneously:
• plan ideas
• select vocabulary
DOI: https://doi.org/10.71112/xy8ca073
717 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
• apply grammar
• monitor pronunciation
• maintain coherence
• manage anxiety
For beginners, this cognitive load is overwhelming. Gemini reduces cognitive
load by providing:
• suggested structures
• corrected versions
• sentence models
• guided questions
When the cognitive load decreases, learners can focus more on coherence and fluency.
Over time, repeated supported output becomes automated. This leads to measurable
improvements in speaking performance. Thus, Gemini is not only a source of input. It is a tool
that stimulates output, supports interaction, and contributes to fluency development through
cognitive optimization (Rahman, 2025).
AFFECTIVE FILTER, ANXIETY, AND CONFIDENCE IN SPEAKING
Many EFL students experience: fear of mistakes, peer judgment, teacher evaluation,
embarrassment, low self-confidence. This is explained by the Affective Filter Hypothesis, which
proposes that anxiety blocks language acquisition. When learners feel threatened, their
cognitive resources shift from language processing to self-protection (Wang, 2020). Gemini
provides what can be described as a “safe-to-fail environment.” The AI does not judge, laugh, or
punish. Learners can repeat attempts without shame. This reduces anxiety and increases
willingness to communicate.
Confidence is not a minor variable. It is often the first condition for improvement. When
learners feel safe, they speak more. When they speak more, they improve. This creates a
DOI: https://doi.org/10.71112/xy8ca073
718 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
positive learning cycle (Novia, 2023). This theoretical point strongly supports your conclusion:
Gemini optimizes both:
• cognitive mechanisms (noticing, output, automation)
• emotional mechanisms (confidence, reduced anxiety)
This synergy is precisely what makes AI tools valuable in speaking development.
ETHICAL AND PEDAGOGICAL CHALLENGES IN AI INTEGRATION
Despite its benefits, the integration of AI raises serious ethical and pedagogical concerns
such as:
Hallucinations and misinformation: AI systems may generate incorrect grammar
explanations or unnatural phrases. This requires teacher supervision and learner training in
verification.
Overdependence: students may become passive and rely on AI instead of developing
independent competence.
Privacy: speaking practice may involve voice, personal data, and sensitive information.
Academic integrity: AI can be used for shortcuts rather than learning. Therefore, the
pedagogical framing is crucial: Gemini must be positioned as a practice partner, not a
replacement for student performance. The teacher’s role becomes central in ethical guidance
and in designing tasks that promote learning rather than dependency (Miguez, 2025).
METODOLOGY
This research adopted a qualitative-interpretive approach with a cross-sectional design
to evaluate the implementation of Artificial Intelligence (AI) at the Instituto Superior Sucre during
the 2025-1st period. From a constructivist perspective, this study sought to understand the
meanings individuals attribute to their experiences in virtual environments, prioritizing
'descriptive richness' and the exploration of complex phenomena such as social perceptions
DOI: https://doi.org/10.71112/xy8ca073
719 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
and behaviors toward technology. To this end, theoretical methods such as analysis-synthesis
approaches were employed, providing a conceptual foundation for the evolution of AI in
education. These were complemented by empirical methods, including structured observation in
Zoom (based on Tamayo's framework) and the application of surveys with online formularies.
This methodological combination enabled the researcher to understand the influence of AI,
particularly in language learning, by integrating narrative and descriptive data that contrasted
student perceptions with actual classroom practices, thereby ensuring a comprehensive view of
results (Bile, 2025).
The population was composed of 15 students of Sucre institute at a2 level who
participated in the observation process from all programs at the Sucre Institute, including
electromechanics, marketing, initial education, electronic, food procedures, audiovisual
production, all of them technical careers. The entire population was contacted via online though
zoom sessions, with responses from 15 students. This represents a convenience sample
selected due to the accessibility and willingness of the participants. This sampling method is
common in educational technology studies when seeking an initial representation of student
perceptions (Bond, 2021).
Theoretical and analytical methods were employed to extract the scientific basis of the
study, extracting and sharing it to develop the theoretical framework. This included the
historical-logical method to analyze the evolution of AI in education and its growing expansion
and influence, the systemic method to observe and understand the interactions between the
components of virtual environments and their users, and the analysis-synthesis method to
integrate the relevant literature. These theoretical methods were used exclusively to establish
the conceptual framework of the study, following approaches established in educational
research (Jhon W. Creswell, 2018)). Empirical methods included the application of a survey and
structured observation of virtual classrooms on the Zoom platform.
DOI: https://doi.org/10.71112/xy8ca073
720 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
TECHNIQUES AND INSTRUMENTS
A Google Forms survey with specific questions designed to achieve the objective of this
study was created, following methodologies validated in educational technology research (Bond,
2021). The instrument included multiple-choice questions, ratings, and true/false statements,
designed to assess student satisfaction and the use of AI tools for their activities in virtual
environments. The form was validated through testing with external individuals and faculty,
adjusting initial questions to improve clarity and precision—a standard procedure to ensure the
instrument's reliability (Creswell & Creswell, 2018). The survey was distributed via Zoom to the
15 students, achieving a response rate of 93.3% (14 responses).
A structured observation was conducted in all active virtual classrooms on the Zoom
platform, using a Microsoft Word form adapted from approaches described by Muijs et al. (2018)
for virtual environments. This form recorded three key variables: the number of assigned
activities, the instructor's modification of the template, and the number of resources used. These
variables allowed for the evaluation of usability and activity levels in each classroom, identifying
patterns in the adoption of AI tools. Access to the classrooms was granted using an authorized
administrator account, ensuring a systematic and objective review of all active classrooms.
Survey data were analyzed using descriptive statistics, calculating frequency and
percentages for multiple-choice, rating, and true/false responses, following standard practices in
educational research (Creswell & Creswell, 2018). This approach allowed for the quantification
of student satisfaction and the perceived use of AI tools. Observational data were processed
using thematic analysis, following the guidelines of Xu and Zammit (2020), thematically coding
the Word document records to identify patterns of use and integration of artificial intelligence in
virtual environments adoption of AI tools in virtual classrooms. Thematic coding included
categories such as activity level, AI integration, and content personalization, ensuring a rigorous
interpretation of the qualitative data.
DOI: https://doi.org/10.71112/xy8ca073
721 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
No conflicts of interest were identified in this study. The survey was anonymous, with no
collection of personally identifiable information, and participants were informed that the survey
was anonymous and complied with ethical research principles, the correspond consent was
implied upon completion of the form.
RESULTS
To interpret the results and formulate de discussion, it was necessary to implement a
survey with 3 sections and some questions in each one, a totally of 8 questions were
implemented with a 5-point Likert scale is used. This instrument seeks to operationalize
qualitative variables (perception, confidence, fluency) into quantitative data for subsequent
statistical analysis. The first section shows:
Figure 1
Perception of the interaction
A big support from nine people who expressed acceptance and the rest who endorsed
the premise, it is validated that students not only consume content but also detect gaps in their
knowledge, solidifying Gemini as a highly accurate lexical tutor. This high score suggests that
the AI model influences speech automation by providing recommended structures that reduce
DOI: https://doi.org/10.71112/xy8ca073
722 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
cognitive load, allowing mental effort to shift from grammar to fluency (Fitria, 2025). Under
Box's model, this implies a reduction in "noise" and variability, stabilizing oral production.
Furthermore, when measuring "linguistic anxiety" under Sagan's concept of science as a
beacon against fear, interaction with a non-punitive entity facilitates experimentation in a safe
environment, leading to the conclusion that improved confidence precedes accuracy and
positions Gemini as a critical tool for the communicative warm-up phase (Sagan, 2013). The
data suggest that the integration of Gemini-AI into speaking practice generates a synergistic
effect: while the AI provides the necessary technical input (Noticing), it simultaneously
optimizes the emotional conditions (Confidence) so that the output is more fluid. However,
according to Popper's logic, these perception results must be contrasted in a later phase with
objective performance measurements (such as pre/post intervention recordings) to avoid social
desirability bias in surveys (Nguyen Thi Phuong Nhung, Noticing Hypothesis in Second
Language Acquisition, 2020).
Section II: Acquisition Mechanism and Knowledge Transfer
Figure 2
Acquisition mechanism and knowledge transfer
DOI: https://doi.org/10.71112/xy8ca073
723 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
Section II confirms that the integration of Gemini-AI transcends mere subjective
perception to become an engine for cognitive restructuring and proactive transfer (Bile, 2025).
The technical analysis of Formula Identification demonstrates that the tool is highly effective for
Input Enhancement, allowing students to break down language into functional units and map
complex structures for deep learning. Under the standards of Ecological Validity, the reported
use of these expressions in real-world contexts constitutes the "gold standard" of the research;
this validates the practical utility of the AI, transforming it from a chat interface into an
instructional resource with a tangible impact on communicative performance. Finally, applying
Popper's falsification logic, the predominance of affirmative responses refutes the hypothesis of
ineffectiveness, solidifying Gemini as a high-precision learning environment that meets the
strictest criteria of internal and external validity in contemporary educational research
(Collaguazo, 2025).
SECTION III: ANALYSIS OF DEPTH AND FREQUENCY
Figure 3
Analysis of depth and frequency
DOI: https://doi.org/10.71112/xy8ca073
724 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
In this last section, it breaks down the depth of Gemini's technical impact, positioning it
not merely as a superficial support but as a catalyst for linguistic complexity. By analyzing the
Effectiveness of Generated Texts, it identifies whether the AI acts as a facilitator of discourse
cohesion through connectives or as an essential corrective feedback mechanism to prevent
linguistic fossilization, following the precepts of The Craft of Research. Likewise, measuring the
Frequency of Noticing and Mimicry reveals the student's level of cognitive engagement; a high
frequency in identifying natural phrases demonstrates the tool's ecological validity by bridging
the gap between academic English and real-world, idiomatic usage (Chan, 2005). Finally,
through a Competency Correlation, the AI is subjected to a contrastive test against traditional
instruction; following Popper's logic, Gemini's superiority in modeling complex thought refutes
the reductionist view of LLMs as mere grammar correctors, solidifying them as advanced
instruments for higher cognitive development and effective transfer to everyday speech (Bubas,
2025).
It demonstrates that the integration of Gemini-AI is not merely a superficial support, but
rather a catalyst for linguistic complexity. While traditional teaching often segments learning,
Gemini offers a holistic model that allows students to identify logical connectives and natural
expressions in real time.
Following Sagan's critical thinking, this data allows us to 'separate the wheat from the
chaff': the true value of AI lies not only in correcting a word, but in providing models of complex
thought that students can emulate, achieving effective transfer to their everyday speech. This
finding is fundamental to recommending AI as a structural component in modern language
curricula, under rigorous technical and ethical standards (Aguilar, 2023).
CONCLUSIONS
Integrating Gemini-AI into English teaching creates a "safe-to-fail" environment that
reduces student anxiety and builds the confidence necessary for fluent speaking. Research
DOI: https://doi.org/10.71112/xy8ca073
725 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
shows that AI acts as a high-accuracy tutor, helping students move beyond simple grammar to
master complex phrases and natural communication in real-time. Because it provides
immediate, personalized feedback that traditional classrooms often cannot, Gemini-AI is a
powerful instructional resource for high-level language development and effective knowledge
transfer (Rahmawati, 2023).
Therefore, the adoption of Gemini-AI is recommended not as a peripheral supplement,
but as a robust instructional resource that achieves effective, proactive knowledge transfer,
meeting the most rigorous technical and ethical standards of contemporary educational
research (Collaguazo, 2025).
Declaration of conflict of interest
The author declares no conflict of interest related to this research.
Declaration of authorship contribution
Miguel Miguez Gordillo: research, writing – revision and editing, methodology and
conceptualization.
Artificial intelligence usage statement
The author declares that artificial intelligence tools were used solely as technical support
in the drafting, organization, and improvement of the linguistic style of this scientific article.
These tools did not replace the author's intellectual, analytical, and critical work at any time. The
manuscript was prepared in accordance with the principles of academic integrity.
DOI: https://doi.org/10.71112/xy8ca073
726 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
REFERENCIES
Abbas Ali Rezaee, Z. A. (2012). The role of zone of proximal development in the students’
learning of English adverbs. Journal of Language Teaching and Research, 3(1), 51–
57. https://doi.org/10.4304/jltr.3.1.51-57
Aguilar. (2023). Want to try Google’s new AI chatbot? Here’s how to sign up for
Bard. CNET. https://www.cnet.com/tech/services-and-software/want-to-try-googles-new-
ai-chatbot-heres-how-to-sign-up-for-bard/
Alexandra Harry. (2023, marzo). Role of AI in education. INJURITY: Journal of Interdisciplinary
Studies, 2(3), 2963–3397. https://doi.org/10.58631/injurity.v2i3.52
Benyo, A. (2020). CALL in English language teaching. International Journal of Advanced
Science and Technology, 1390–1395.
Bile, A. (2025, octubre 10). Intelligent learning: AI’s impact on early educational
strategies. https://doi.org/10.1002/9781394352821.ch04
Bond. (2021). Schools and emergency remote education during the COVID-19 pandemic: A
living rapid systematic review. Asian Journal of Distance Education, 15(2), 191–
247. https://doi.org/10.5281/zenodo.4425683
Booth, W. C. (2009). The craft of research (Vol. 1). https://books.google.com.ec
Bubas, G. (2025). How ChatGPT and Gemini view the elements of communication competence
of large language models: A pilot study. arXiv. https://doi.org/10.48550/arXiv.2511.02838
Cabrera, T. (2014, enero 30). El concepto y la visión del desarrollo como base para la
evaluación de políticas públicas. Revista Economía y Sociedad, 20.
Chan, T. P. (2005). Effects of web-based concordancing instruction on EFL. Computer Assisted
Language Learning, 18(3), 231–251. https://doi.org/10.1080/09588220500185769
Collaguazo, J. Z. (2025). The use of Gemini app to enhance oral communication in EFL
classroom: A pedagogical innovation. Revista Científica Multidisciplinar G-Nerando, 2,
18–67. https://doi.org/10.60100/rcmg.v6i2.769
DOI: https://doi.org/10.71112/xy8ca073
727 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
Creswell, J. W., & Creswell, J. D. (2018). Research design. https://books.google.com.ec
Eko S, W. (2025). Improving students’ speaking and listening skills using Gemini AI
application. https://eprints.untirta.ac.id/id/eprint/54533
Fitria, T. N. (2025). Revolutionizing English learning with AI: Insights from ChatGPT and Google
Gemini. International Journal of Computer and Information System, 6(2). https://ijcis.net
Guendouz, S. (2025). The role of artificial intelligence in achieving fluency in oral expression
classes. https://dspace.univ-ouargla.dz
Hasanein, A. M., Sobaih, A. E., & Elshaer, I. A. (2024). Examining Google Gemini’s acceptance
and usage in higher education. Journal of Applied Learning & Teaching, 7(2), 223–
231. https://doi.org/10.3316/informit.T2025091800012891397238008
Hernando Barrios-Tao, V., et al. (2026). Impact of artificial intelligence on educational actors
(2015–2023). Revista Colombiana de Ciencias Sociales, 16(1), 240–
276. https://doi.org/10.21501/22161201.4803
Katikela Kishore, R. S. (2024). Evaluating Telugu proficiency in large language models: A
comparative analysis of ChatGPT and
Gemini. arXiv. https://doi.org/10.48550/arXiv.2404.19369
Krashen, S. (2020). The optimal input hypothesis: Not all comprehensible input is of equal
value. CATESOL Newsletter, 5(1), 1–2.
Long, M. H. (1983). Native speaker/non-native speaker conversation and the negotiation of
comprehensible input. Applied Linguistics, 4(2), 126–
141. https://doi.org/10.1093/applin/4.2.126
Míguez, M. (2025). La inteligencia artificial como recurso didáctico en el proceso de enseñanza
y aprendizaje. Revista Multidisciplinar Epistemología de las Ciencias, 2(2), 181–
196. https://doi.org/10.71112/vqb1zx28
DOI: https://doi.org/10.71112/xy8ca073
728 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
Ng, M. L., et al. (2025). Grammar and engagement in focus: Evaluating Gemini AI’s impact on
an educational environment. Computers and Education Open, 9,
100302. https://doi.org/10.1016/j.caeo.2025.100302
Nguyen Thi Phuong Nhung, M. (2020). Noticing hypothesis in second language
acquisition. IOSR Journal of Humanities and Social Science, 25(1), 26–30.
Novia, F. R. (2023). The correlation between students’ self-confidence and speaking
achievement. Journal of English Education, 4(1), 35–
43. https://doi.org/10.52333/djoee.v4i1.62
Omar Mahmoud ElSenbawy, K. B. (2025). Use of generative large language models for patient
education on common surgical conditions: A comparative analysis between ChatGPT
and Google Gemini. Updates in Surgery, 78, 469–475. https://doi.org/10.1007/s13304-
025-02074-8
Purnama, M. R. (2025). Applying Gemini to enhance EFL learning: A review of advantages and
challenges. Celtic: A Journal of Culture, English Language Teaching, Literature and
Linguistics, 12(2), 990–1004. https://doi.org/10.22219/celtic.v12i2.42739
Rahman, M. M. (2025). Reducing anxiety and enhancing motivation: Gemini AI voice chat for
ESL. Journal of Linguistics & Language Studies, 1(2), 56–87.
Rahmawati, A. S. (2023). Improving speaking skills through TikTok application. Journal of
Languages and Language Teaching, 11(1), 137. https://doi.org/10.33394/jollt.v11i1.6633
Rashtchi, M., & Ebrahimi, F. (2018). Learning the English passive voice: A comparative study
on input flooding and input enhancement techniques. International Linguistics Research,
1(1), 67. https://doi.org/10.30560/ilr.v1n1p67
Sagan, D. (2013). Cosmic apprentice: Dispatches from the edges of
science. https://books.google.com.ec
DOI: https://doi.org/10.71112/xy8ca073
729 Multidisciplinary Journal Epistemology of the Sciences | Vol. 3, Issue 2, 2026, April–June
Siti Nurul Azizah, E. H. (2025). Gemini AI as a writing catalyst: Boosting fluency, coherence,
and confidence in EFL composition. Journal of English Education, 3(2), 64–
77. https://doi.org/10.61994/jee.v3i2.1134
Swain, M. (2005). The output hypothesis. https://www.taylorfrancis.com
Wang, L. (2020). Application of affective filter hypothesis in junior English vocabulary
teaching. Journal of Language Teaching and Research, 11(6), 983–
987. http://dx.doi.org/10.17507/jltr.1106.16
Zammit. (2020). The incubation period during the pandemic of COVID-19: A systematic review
and meta-analysis. Systematic Reviews, 10, 101. https://doi.org/10.1186/s13643-021-
01648-y
