CoChat
Sign Up
Cited by 17Second Language Acquisition and LearningArtificial Intelligence in EducationEnglish Language Learning and Teaching

Integrating ChatGPT for vocabulary learning and retention: A classroom-based study of Saudi EFL learners

Open access
Abdelhalim, Safaa Mahmoud, Alsehibany, Raniya·2025·Language Learning & Technology
DOI 10.64152/10125/73635
Open this paper in CoChat — free accountOpen PDF

Abstract

Grounded in Vygotsky's sociocultural theory, this study examines the effectiveness of integrating ChatGPT into classroom instruction to enhance vocabulary learning. It evaluates ChatGPT's impact on vocabulary knowledge and retention compared to traditional teacher-led instruction. Using a quasi-experimental pre- and posttest control group design within an explanatory sequential mixed-methods approach, 71 EFL learners were randomly assigned to an experimental group (EG) and a control group (CG). The EG used ChatGPT as an interactive language tool, engaging in vocabulary practice, receiving real-time feedback, and completing AI-generated exercises tailored to their progress, while the CG followed traditional lecture-based instruction. Data collection included vocabulary knowledge tests (pretest, posttest, and delayed posttest) and semi-structured interviews. Quantitative results showed that the EG significantly outperformed the CG in productive vocabulary knowledge and overall test scores, with moderate to large effect sizes. While receptive vocabulary gains did not differ significantly between groups, within-group comparisons confirmed substantial improvements. Qualitative findings highlighted students' positive perceptions, emphasizing ChatGPT's role in enhancing engagement, motivation, and contextual vocabulary use through scaffolded interaction. This study contributes to the growing field of AI-driven language learning by demonstrating how structured, theory-informed integration of ChatGPT can effectively support productive vocabulary development, offering both practical and pedagogical insights.

AI snapshot

Written from the paper’s full text.

TL;DR

This quasi-experimental mixed-methods study with 71 Saudi female EFL learners found that classroom integration of ChatGPT improved productive vocabulary knowledge and overall vocabulary scores compared with traditional instruction, but not receptive vocabulary between groups. Qualitative interviews showed positive perceptions, with students emphasizing engagement, feedback, and contextual vocabulary use under teacher scaffolding.

Key findings

  • On the posttest, the experimental group significantly outperformed the control group in productive vocabulary knowledge (U = 401.50, Z = 2.64, p = .008, r = .36) and overall vocabulary scores (U = 405.50, Z = 2.59, p = .010, r = .36), but not in receptive knowledge (U = 532.00, Z = 1.13, p = .257).
  • Within the experimental group, pre-to-post gains were significant for receptive knowledge (Z = 2.16, p = .031, r = .46), productive knowledge (Z = 4.86, p = .000, r = .94), and overall scores (Z = 4.89, p = .000, r = .95).
  • On the delayed posttest, the experimental group maintained significantly higher productive vocabulary (U = 437.50, Z = 2.23, p = .026, r = .31) and overall scores (U = 438.50, Z = 2.21, p = .027, r = .30), with no significant receptive difference (U = 502.50, Z = 1.47, p = .141).
  • Thematic analysis produced four themes: ease of use and initial adaptation; enhanced vocabulary acquisition and retention; effective feedback, interaction, and usability; and increased engagement, confidence, and extended learning opportunities.

Methods

  • Quasi-experimental pre/post/delayed posttest control-group design within an explanatory sequential mixed-methods approach; 71 Arabic L1 female sophomores (ages 19–21, M = 19.99, SD = 1.44), all A2-level, were randomly assigned to a control group (n = 35) or experimental group (n = 36) at a Saudi technical college, taught by the same instructor.
  • The experimental group used ChatGPT in six weeks of 50-minute weekly lessons for vocabulary practice, real-time feedback, and AI-generated exercises with scaffolded prompts; the control group followed traditional lecture-based instruction and was instructed not to use AI tools. Vocabulary tests included 20 multiple-choice receptive items covering 29 target words and 19 productive sentence-construction items; Cronbach’s alpha was .79 overall, .81 receptive, and .79 productive. Six experimental-group students were interviewed.
  • Quantitative data were analyzed in SPSS v28; Shapiro-Wilk indicated non-normal distributions, so Wilcoxon Signed Ranks tests were used within groups, Mann-Whitney U tests between groups, and Pearson’s r for effect sizes. Qualitative interviews were analyzed using thematic analysis, with inter-rater reliability reported as r = .93.

Limitations

  • Stated limitations: the study was conducted in a single institutional setting with a relatively small, homogeneous sample, limiting generalizability; it primarily focused on short-term effects, and the four-week delayed posttest may not capture long-term learning; vocabulary was assessed in terms of meaning, form, and use without addressing collocations, associations, or depth of knowledge.
  • Stated future needs: research should include larger and more varied groups, male and mixed-gender groups, diverse educational contexts, longer follow-ups, broader vocabulary constructs, and learners’ AI literacy, critical thinking, and ethical awareness.
  • Evident from the text: the sample was all female Arabic L1 A2 sophomores at one technical college, so gender, proficiency, and context are constrained; outside-class ChatGPT use by the experimental group was optional and not graded, and control-group compliance was monitored through verbal check-ins and teacher logs, which may leave some variation in exposure unmeasured.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.64152/10125/73635

Cites

61 of 65 references verified0 citation errors · 4 not checkable

Statistics

Claims the paper makes

24 of 47 cited sources checked · 8 supported · 1 flagged · accuracy not yet measured

  • Premisecitation

    Saudi EFL learners have been found to favor passive learning strategies over more cognitively demanding techniques.

    “Saudi EFL learners have been found to favor passive learning strategies (e.g., rote memorization and repetition) over more cognitively demanding techniques (Ali, 2020;Alshammari, 2020).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Ali, 2020 source
    • cites Alshammari, 2020 source

      The Discussion (page unavailable) reports that Saudi students with larger vocabularies used all strategy types equally, while those with smaller vocabularies used metacognitive strategies most. It attributes a preference for repetition over deeper cognitive strategies to studies of Chinese and Australian learners, not to its Saudi participants. (LLM, conf=0.88)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
  • Interpretationown data

    Our findings contrast with earlier research which found that technology-based vocabulary instruction primarily benefited receptive knowledge rather than production.

    “Interestingly, our findings contrast with earlier research (e.g., Allal-Sumoto et al., 2024;Bueno-Alastuey & Nemeth, 2020), which found that technology-based vocabulary instruction primarily benefited receptive knowledge rather than production.”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Allal-Sumoto et al., 2024 source
    • cites Bueno-Alastuey & Nemeth, 2020 source
  • Interpretationcitation

    ChatGPT's ability to deliver immediate feedback and contextualized examples contributed to learners' vocabulary gains.

    “ChatGPT's ability to deliver immediate feedback and contextualized examples appears to have contributed to learners' vocabulary gains (Karataş et al., 2024;Liu et al., 2024).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Karataş et al., 2024 source
    • cites Liu et al., 2024 source

      The Introduction says ChatGPT can provide instant feedback, and “Actual use of chatbots as tutors or conversation partners” reports learners using AI to create vocabulary games for learning new words. The paper also describes contextualized answers, but these passages do not link feedback or contextualized examples to vocabulary gains; its reported feedback example concerns grammar. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
  • Premisecitation

    AI-based tools function as dynamic mediators, continuously assessing learner input and offering ZPD-sensitive feedback, guidance, and modeling.

    “In recent research, AI-based tools have been shown to function as dynamic mediators, continuously assessing learner input and offering ZPD-sensitive feedback, guidance, and modeling (Mingyan et al., 2025;Wei, 2023).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Mingyan et al., 2025

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Wei, 2023 source

      The Theoretical framework describes the AI tool as a collaborative partner helping learners advance toward their ZPD, and the Discussion reports adaptive support, feedback tailored to learners’ written texts, guidance, and alternative sentences. The paper does not establish that the tool continuously assessed learner input or calibrated its feedback specifically to each learner’s ZPD. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Premisecitation

    Learners need multiple contextualized exposures to internalize vocabulary effectively.

    “Nation (2015) emphasized that learners need multiple contextualized exposures to internalize vocabulary effectively.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Nation (2015) source
  • May combine two claims. 40
  • Classified as probably not a substantial claim. 25

  • May combine two claims.

    Premisecitation

    ChatGPT supports vocabulary acquisition by presenting language in authentic, meaningful contexts and offering cultural insights, which can enhance retention and learner engagement.

    “ChatGPT also supports vocabulary acquisition by presenting language in authentic, meaningful contexts and offering cultural insights, which can enhance retention and learner engagement (Karataş et al., 2024;Kohnke et al., 2023).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Karataş et al., 2024 source

      In §3.1, a student says ChatGPT helped them learn words in context and remember them, and participants report greater engagement and opportunities to explore cultures. Section 4.4 says its materials can offer cultural insights, but recommends that teachers ensure exposure to authentic, real-life communicative practice; it does not establish that ChatGPT itself provides that practice. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites Kohnke et al., 2023 source

  • May combine two claims.

    Premisecitation

    English instruction in Saudi Arabia remains largely teacher-centered, emphasizing rote memorization over interactive learning.

    “English instruction remains largely teacher-centered, emphasizing rote memorization over interactive learning (Alshammari, 2020).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Alshammari, 2020 source

      The Introduction and ‘Vocabulary learning in the Saudi context’ section discuss limited progress and a gap between learners’ needs and school pedagogy, but do not describe English instruction as teacher-centered or reliant on rote memorization. The Abstract reports that lower-achieving students focus on memory strategies; that is a finding about learners, not teaching practice. Searches for teacher-centered instruction, rote memorization, and interactive learning found no passage establishing the claimed contrast. (LLM, conf=0.78)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Saudi EFL learners struggle with vocabulary retention and application in writing and speaking.

    “Consequently, learners struggle with vocabulary retention and application in writing and speaking (Ababneh, 2017;Alsehibany & Abdelhalim, 2023), revealing a gap between curricular objectives and actual proficiency (Al-Seghayer, 2017).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Ababneh, 2017 source

      The Results and Discussions report improper word choices in the writing of the female Saudi university students studied, supporting difficulty applying vocabulary in writing. The paper does not assess vocabulary retention or speaking, and it cautions that its findings cannot be generalized to all Saudi learners. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Alsehibany & Abdelhalim, 2023 source

  • May combine two claims.

    Premisecitation

    ChatGPT is not confined to pre-programmed responses and adapts flexibly to learner input, making it well-suited for vocabulary development through dynamic and personalized interactions.

    “Unlike many purpose-built AI language applications, ChatGPT is not confined to pre-programmed responses; instead, it adapts flexibly to learner input, making it particularly well-suited for promoting vocabulary development through dynamic and personalized interactions (Karataş et al., 2024).”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Karataş et al., 2024 source

      Section 4.5 says ChatGPT tailors responses to individual prompts and offers interactive, personalized feedback; section 3.1 reports perceived benefits for new vocabulary acquisition. The paper does not establish the claimed contrast with purpose-built language applications or show that their responses are pre-programmed. (LLM, conf=0.85)

      Experimental — accuracy not yet measured

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    ChatGPT is open accessible, has extensive linguistic capabilities, and can simulate human-like dialogue across topics, providing immersive learning experiences that foster deeper cognitive processing.

    “Among these tools, ChatGPT has garnered significant attention due to its open accessibility, extensive linguistic capabilities, and ability to simulate human-like dialogue across a wide range of topics, thereby providing immersive learning experiences that foster deeper cognitive processing (Mugableh, 2024;Zhang & Huang, 2024).”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites Mugableh, 2024 source
    • cites Zhang & Huang, 2024 source

      The Introduction (page not provided) says AI chatbots mimic human conversation, and the Conclusion (page not provided) says LLM chatbots generate natural, fluent responses and provide an interactive language-learning environment. The paper supports the dialogue and linguistic-capability aspects at the level of LLM chatbots, but it does not establish ChatGPT’s open accessibility or show that immersive experiences foster deeper cognitive processing. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Saudi EFL learners face difficulties in vocabulary acquisition due to limited English exposure outside the classroom and linguistic differences between Arabic and English.

    “Saudi EFL learners face difficulties in vocabulary acquisition due to limited English exposure outside the classroom and linguistic differences between Arabic and English (Afzal, 2019;Altalhab, 2019).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Afzal, 2019 source

      The Abstract (p. 1) reports vocabulary-learning difficulties among BA English majors at a Saudi university. It does not attribute those difficulties to limited English exposure outside the classroom or differences between Arabic and English, and its findings concern those students rather than Saudi EFL learners generally (searched: exposure, Arabic). (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites Altalhab, 2019 source

      The Abstract and Conclusion report limited vocabulary knowledge among the Saudi tertiary students studied, supporting the claim that they face vocabulary difficulties. The paper does not attribute those difficulties to limited English exposure outside the classroom or to linguistic differences between Arabic and English. Results and Discussion instead suggests that English-language technology may have increased exposure. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The results align with prior research advocating GenAI chatbots in language learning, particularly for promoting interactive, adaptive engagement.

    “These results align with prior research advocating GenAI chatbots in language learning, particularly for promoting interactive, adaptive engagement (Kohnke et al., 2023;Yigci et al., 2024;Zhang & Huang, 2024).”
    Quote in the paperNot a substantial claim2 of 3 sources checked
    • cites Kohnke et al., 2023 source
    • cites Yigci et al., 2024 source
    • cites Zhang & Huang, 2024 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The results can be interpreted through Vygotsky's sociocultural theory, specifically the ZPD.

    “This can be interpreted through Vygotsky's (1978) sociocultural theory, specifically the ZPD, which emphasizes scaffolded support to help learners advance beyond their current competence (Lantolf, 2010;van Compernolle & Williams, 2013).”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Vygotsky's (1978)

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Lantolf, 2010 source
    • cites van Compernolle & Williams, 2013 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Zhang and Huang (2024) found that an LLM-powered chatbot led to gains in both receptive and productive vocabulary among Chinese high school students.

    “Zhang and Huang (2024) investigated an LLM-powered chatbot's role in vocabulary learning among Chinese high school students, reporting gains in both receptive and productive vocabulary.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Zhang and Huang (2024) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    ChatGPT may not be culturally sensitive, resulting in instances that are inconsistent with learners' sociolinguistic backgrounds.

    “Additionally, ChatGPT may not be culturally sensitive, resulting in instances that are inconsistent with learners' sociolinguistic backgrounds (Crompton & Burke, 2024).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Crompton & Burke, 2024 source

      The Potential Bias section (page unavailable) says ChatGPT can reflect cultural biases in its training data, supporting the concern about cultural sensitivity. It does not say that ChatGPT produces responses inconsistent with learners’ sociolinguistic backgrounds. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationown data

    These findings support previous research on technology as a mediator of vocabulary learning and learner autonomy.

    “These findings support previous research on technology as a mediator of vocabulary learning and learner autonomy (Bashori et al., 2022;Lai & Chen, 2023;Teng, 2022).”
    Quote in the paperNot a substantial claim2 of 3 sources checked
    • cites Bashori et al., 2022 source

      Section 4.1 reports that students using the ASR-based websites significantly outperformed controls on two receptive-vocabulary tests, supporting the technology-and-vocabulary part of the claim. The Discussion describes personalized training and feedback, but our copy does not establish that the websites improved learner autonomy; the References list a separate work on pronunciation learner autonomy. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites Lai & Chen, 2023

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Teng, 2022 source

      The concluding remarks (p. 29) report that captioned videos, especially glossed full captions, support L2 vocabulary learning. The paper does not report a finding about learner autonomy; that term appears in the title of a different work listed in the References (p. 33). (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Studies on ChatGPT in language learning found that AI-powered chatbots facilitate contextualized vocabulary acquisition, exposing learners to authentic, real-world language use.

    “Similarly, studies on ChatGPT in language learning (e.g., Karataş et al., 2024;Zhang & Huang, 2024) found that AI-powered chatbots facilitate contextualized vocabulary acquisition, exposing learners to authentic, real-world language use.”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites Karataş et al., 2024 source

      Section 3.1 reports that a student learned new words in context with ChatGPT, supporting contextualized vocabulary acquisition. But section 4.4 recommends ensuring students also receive authentic, real-life communicative practice; it does not establish that ChatGPT itself exposes learners to authentic, real-world language use. Page numbers were not provided in our copy. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Zhang & Huang, 2024 source

      The Finding and discussion section reports improved vocabulary acquisition and incidental learning with LLM chatbots. The Literature review describes learning words through context, and the Methodology describes examples and conversation practice. The paper does not establish that learners were exposed to authentic, real-world language use; its described activities include chatbot conversations and mock interviews (page numbers unavailable). (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Findingown data

    Students initially encountered challenges with ChatGPT, but with targeted teacher guidance and practice, the tool became accessible and user-friendly.

    “Students initially encountered challenges with ChatGPT; however, with targeted teacher guidanceespecially training on writing prompts, provision of examples during the orientation session, and collaborative classroom practice-the tool quickly became accessible and user-friendly.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The experimental group significantly outperformed the control group in productive vocabulary knowledge on the posttest, with a moderate effect size.

    “Results (Table 1) show that EG significantly outperformed CG in productive knowledge (U = 401.50,Z = 2.64, p = .008,r = .36)and overall vocabulary scores (U = 405.50,Z = 2.59, p = .010,r = .36),both with moderate effect sizes.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The experimental group's optional outside-class ChatGPT use did not confound the comparison between groups.
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

  • May combine two claims.

    Findingreasoning

    Students had an overall positive perception of integrating ChatGPT into their vocabulary learning process, contrasting it favorably with traditional instructional methods, particularly during structured class activities.

    “Students had an overall positive perception of integrating ChatGPT into their vocabulary learning process, contrasting it favorably with traditional instructional methods, particularly when used during structured class activities.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Within the experimental group, receptive vocabulary knowledge improved significantly from pretest to posttest, with a moderate effect size.

    “As illustrated in Table 2, significant improvements were observed in both receptive (Z = 2.16, p = .031,r = .46)and productive knowledge (Z = 4.86, p = .000,r = .94),with effect sizes ranging from moderate to large.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

  • May combine two claims.

    Findingown data

    The experimental group significantly outperformed the control group in overall vocabulary scores on the posttest, with a moderate effect size.

    “Results (Table 1) show that EG significantly outperformed CG in productive knowledge (U = 401.50,Z = 2.64, p = .008,r = .36)and overall vocabulary scores (U = 405.50,Z = 2.59, p = .010,r = .36),both with moderate effect sizes.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The experimental group's optional outside-class ChatGPT use did not confound the comparison between groups.
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

  • Classified as probably not a substantial claim.

    Findingown data

    Contextual learning and interactive prompts helped students remember and apply words more effectively.

    “Several students emphasized that contextual learning and interactive prompts helped them remember and apply words more effectively.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The experimental group maintained significantly higher overall vocabulary knowledge than the control group on the delayed posttest, with a moderate effect size.

    “EG maintained significantly higher scores in productive knowledge (U = 437.50,Z = 2.23, p = .026,r = .31)and overall vocabulary knowledge (U = 438.50,Z = 2.21, p = .027,r = .30),both with moderate effect sizes.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The experimental group's optional outside-class ChatGPT use did not confound the comparison between groups.
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.
    • Rests on an assumption: The four-week delayed posttest adequately measures long-term retention.

  • May combine two claims.

    Findingown data

    Within the experimental group, overall vocabulary test scores increased significantly from pretest to posttest, with a large effect size.

    “Overall test scores also showed a substantial increase (Z = 4.89, p = .000,r = .95),reinforcing ChatGPT's effectiveness in enhancing vocabulary acquisition.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

  • May combine two claims.

    Findingown data

    Within the experimental group, productive vocabulary knowledge improved significantly from pretest to posttest, with a large effect size.

    “As illustrated in Table 2, significant improvements were observed in both receptive (Z = 2.16, p = .031,r = .46)and productive knowledge (Z = 4.86, p = .000,r = .94),with effect sizes ranging from moderate to large.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

  • May combine two claims.

    Findingreasoning

    ChatGPT contributed to greater learner confidence and encouraged vocabulary use in real contexts.

    “ChatGPT contributed to greater learner confidence and encouraged vocabulary use in real contexts.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingown data

    ChatGPT's integrated examples enhanced students' understanding of vocabulary usage within sentences, conversations, or narratives, leading to improved retention.

    “emphasizing how ChatGPT's integrated examples enhanced their understanding of vocabulary usage within sentences, conversations, or narratives, leading to improved retention.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Some students initially struggled with phrasing effective prompts or tended to over-rely on the tool's output without critically engaging with it.

    “Still, a few students mentioned that they initially struggled with phrasing effective prompts or tended to over-rely on the tool's output without critically engaging with it.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Early challenges highlighted a learning curve in developing prompt literacy and self-regulated use, which required teacher support and reflective classroom practice.

    “These early challenges highlighted a learning curve in developing prompt literacy and self-regulated use, which required both teacher support and reflective classroom practice to overcome.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Students viewed ChatGPT as an engaging and effective tool.

    “Students viewed ChatGPT as an engaging and effective tool, consistent with previous studies linking GenAI use to increased motivation and learner interaction (Liu & Ma, 2023;Xiao & Zhi, 2023).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Students reported that ChatGPT expands vocabulary and enhances ability to form sentences.

    “It expands my vocabulary and enhances my ability to form sentences,”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S1 (asserted), S4 (based).

    Findingown data

    On the delayed posttest, the experimental group significantly outperformed the control group in productive vocabulary knowledge.

    “Productive | Control | 35 | 30.50 | 1067.50 437.50 2.233 .026 | Sig. at | .31”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing), S4 (based).

    Findingown data

    On the delayed posttest, there was no significant difference between the experimental and control groups in receptive vocabulary knowledge.

    “Receptive | Control | 35 | 32.36 | 1132.50 502.50 1.473 .141 | Not Sig -”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Qualitative data revealed students' positive perceptions, highlighting ChatGPT's role in enhancing engagement, confidence, and autonomous learning.

    “Additionally, qualitative data revealed students' positive perceptions, highlighting ChatGPT's role in enhancing engagement, confidence, and autonomous learning.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The interviews were audio-recorded, transcribed verbatim, and subjected to member checking to ensure accuracy and credibility.

    “The interviews were audio-recorded, transcribed verbatim, and subjected to member checking to ensure accuracy and credibility.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductcitation

    English proficiency was assessed using the Oxford Placement Test, and all participants were classified as A2-level.

    “English proficiency was assessed using the Oxford Placement Test (Allen, 2004), chosen for its alignment with CEFR levels and its proven reliability in EFL research.All participants were classified as A2-level-suitable for engaging with AI tools while still requiring vocabulary support.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Allen, 2004

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Compliance with the no-AI instruction in the control group was monitored through regular verbal check-ins and teacher logs.

    “Compliance was monitored through regular verbal check-ins and teacher logs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductown data

    Two independent researchers coded the qualitative data separately and achieved high inter-rater reliability (r = .93).

    “To enhance reliability, two independent researchers coded the data separately, achieving a high inter-rater reliability score (r = .93).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisenone given

    Most existing research on GenAI for vocabulary instruction has examined informal or self-directed learning contexts, leaving a gap in understanding how ChatGPT can be systematically integrated into structured classroom instruction.

    “Moreover, most of the existing research has examined informal or self-directed learning contexts, leaving a gap in understanding how ChatGPT can be systematically integrated into structured classroom instruction.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The receptive vocabulary dimension may require more passive exposure and reading-based strategies, which were less prominent in the intervention.

    “In contrast, the receptive dimension may require more passive exposure and reading-based strategies, which were less prominent in the intervention.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductown data

    Baseline equivalence was confirmed: there were no significant differences between the experimental and control groups in overall vocabulary scores, receptive knowledge, or productive knowledge at pretest.

    “Results showed no significant differences between the EG and CG in overall vocabulary scores (U = 617.00,Z = .17,p = .881)or the receptive (U = 615.00,Z = .17,p = .862)and productive knowledge sub-sections (U = 590.50,Z = .47,p = .639).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The same instructor taught both groups to control instructor bias.

    “The same instructor taught both groups to control instructor bias.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    Learners require teacher facilitation to maximize GenAI's benefits and critically evaluate its outputs.

    “This supports research suggesting that GenAI tools should complement rather than replace human instruction (Arifin et al., 2024), as learners still require teacher facilitation to maximize GenAI's benefits and critically evaluate its outputs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    ChatGPT can simulate dialogic learning environments, facilitate contextualized language use, and promote collaborative meaning-making.

    “this study leverages ChatGPT to simulate dialogic learning environments, facilitate contextualized language use, and promote collaborative meaning-making.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    A dual scaffolding model in which both human and AI support optimized vocabulary learning and fostered learner autonomy.

    “Importantly, students also stressed the need for teacher mediation to validate AI outputs, highlighting a dual scaffolding model in which both human and AI support optimized vocabulary learning and fostered learner autonomy.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The findings align with sociocultural perspectives on learning, particularly Vygotsky's concept of mediation and scaffolding.

    “These findings further align with sociocultural perspectives on learning, particularly Vygotsky's (1978) concept of mediation and scaffolding, which emphasizes the role of structured support in language development (Schumm, 2006).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Vygotsky's (1978)

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Schumm, 2006 source

  • Classified as probably not a substantial claim.

    Interpretationcitation

    The discrepancy may stem from traditional L2 learning challenges, where students often prioritize memorizing word meanings over active use.

    “This discrepancy may stem from traditional L2 learning challenges, where students often prioritize memorizing word meanings over active use (Laufer & Hulstijn, 2001).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Laufer & Hulstijn, 2001 source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Earlier studies suggested that L2 learners prioritize receptive vocabulary knowledge over productive use, focusing on input-based strategies rather than output-driven approaches.

    “These studies suggested that L2 learners prioritize receptive vocabulary knowledge over productive use, focusing on input-based strategies rather than output-driven approaches.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Vocabulary learning depends on declarative memory, strengthened through explicit instruction and repeated exposure.

    “Ullman's (2001) model of second language acquisition suggests vocabulary learning depends on declarative memory, strengthened through explicit instruction and repeated exposure.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Ullman's (2001) source

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Embedding GenAI-enhanced mediation alongside human mediation in structured classroom interactions reinforces sociocultural theory's emphasis on socially constructed knowledge and scaffolded learning.

    “By embedding GenAI-enhanced mediation alongside human mediation (teachers and peers) in structured classroom interactions, this study reinforces the sociocultural theory's emphasis on socially constructed knowledge and scaffolded learning.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Explicit vocabulary instruction is more effective than incidental learning, as structured teaching enhances retention.

    “Explicit vocabulary instruction is widely recognized as more effective than incidental learning, as structured teaching enhances retention (Vincy, 2020).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Vincy, 2020

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationcitation

    GenAI tools should complement rather than replace human instruction.

    “This supports research suggesting that GenAI tools should complement rather than replace human instruction (Arifin et al., 2024), as learners still require teacher facilitation to maximize GenAI's benefits and critically evaluate its outputs.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Arifin et al., 2024 source

  • May combine two claims.

    Interpretationreasoning

    GenAI-powered tools can enhance engagement and facilitate deeper cognitive processing in vocabulary learning while maintaining the essential role of human educators.

    “The integration of AI-driven scaffolding into classroom instruction suggests that GenAI-powered tools can enhance engagement and facilitate deeper cognitive processing in vocabulary learning while maintaining the essential role of human educators in guiding and contextualizing the learning process.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    ChatGPT's ability to provide contextualized examples, adaptive feedback, and interactive engagement mirrors Vygotsky's notion that learners progress most effectively when supported by structured external guidance before transitioning to independent mastery.

    “ChatGPT's ability to provide contextualized examples, adaptive feedback, and interactive engagement mirrors Vygotsky's notion that learners progress most effectively when supported by structured external guidance before transitioning to independent mastery.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    There were no statistically significant pre-test differences between the control and experimental groups in receptive knowledge, productive knowledge, or overall test scores.

    “Receptive Knowledge Control 35 35.571245.00 615.000 .174.862Not Sig.Experimental 36 36.42 1311.00Productive knowledge Control 35 34.87 1220.50 590.500 .470.639Not Sig.Experimental 36 37.10 1335.50Overall test scores Control 35 35.63 1247.00 617.000 .150.881Not Sig.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    Over-reliance on AI, uncritical acceptance of generated content, and exposure to potentially inaccurate or culturally inappropriate examples underscores the need for AI literacy.

    “Over-reliance on AI, uncritical acceptance of generated content, and exposure to potentially inaccurate or culturally inappropriate examples underscores the need for AI literacy.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    ChatGPT's interactive, output-focused approach encouraged students to apply vocabulary in meaningful contexts, resulting in stronger gains in productive knowledge and fostering active vocabulary usage.

    “By contrast, ChatGPT's interactive, output-focused approach encouraged students to apply vocabulary in meaningful contexts, resulting in stronger gains in productive knowledge and fostering active vocabulary usage.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The study primarily focused on short-term effects; the four-week delayed posttest may not capture long-term learning outcomes.

    “The study primarily focused on short-term effects; although a four-week delayed posttest offered preliminary insights into retention, it may not capture long-term learning outcomes.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    Unless properly led, over-reliance on AI-generated content might limit opportunities for deeper language processing or critical thinking.

    “Unless properly led, this might limit opportunities for deeper language processing or critical thinking (Liu & Ma, 2023;Yan, 2023).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Liu & Ma, 2023 source
    • cites Yan, 2023 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Vocabulary knowledge was assessed in terms of meaning, form, and use without addressing other aspects such as collocations, associations, or depth of knowledge.

    “Additionally, vocabulary knowledge was assessed in terms of meaning, form, and use without addressing other important aspects such as collocations, associations, or depth of knowledge.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    There is a risk of over-relying on AI-generated content.

    “According to recent research and students' feedback, there is a risk of over-relying on AI-generated content.”
    Quote in the paperNot a substantial claim
57 of 57

  • The paper relies on this without support.

    The experimental group's optional outside-class ChatGPT use did not confound the comparison between groups.

    “While EG students were encouraged to use ChatGPT outside class for additional practice, this was strictly optional and not tied to the assessment.”
    “To reduce external validity threats, these tasks were clearly marked as optional and not graded.”

  • The paper relies on this without support.

    The vocabulary knowledge test validly measures receptive and productive vocabulary knowledge.

    “The vocabulary knowledge test functioned as both an initial and final assessment, enabling the measurement of students' progress and retention of vocabulary over time.”
    “It was divided into two sections to comprehensively assess both receptive and productive vocabulary knowledge.”

  • The paper relies on this without support.

    The four-week delayed posttest adequately measures long-term retention.

    “Four weeks later, a delayed posttest was administered to assess retention.”
3 of 3

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • AI-mediated interactions may support speaking development and vocabulary learning through scaffolded dialogue and adaptive feedback.

    “emerging research suggests that AI-mediated interactions may support speaking development and vocabulary learning through scaffolded dialogue and adaptive feedback (Fathi et al., 2024; Abdelhalim and Alsehibany, 2025).”

    From Perceptions vs. practices: academic integrity and actual ChatGPT use among EFL students (2026) · cites this paper as Abdelhalim and Alsehibany, 2025

    The Discussion describes AI-driven scaffolding, adaptive feedback, and interactive engagement as supporting vocabulary learning. The Given section mentions speaking only as a potential application through conversation simulations; it does not report speaking-development outcomes. Thus, the source supports the vocabulary portion of the claim but not speaking development as a study finding. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • This review conceptualizes fluency illusion as a key explanatory mechanism shaping learners' experiences with fluent external supports.

    “By integrating insights from processing fluency and metacognition, this review conceptualizes fluency illusion as a key explanatory mechanism shaping learners' experiences with fluent external supports [32].”

    From Fluency Illusion: A Review on Influence of ChatGPT in Classroom Settings (2026) · cites this paper as [32]

    The Effective Feedback, Interaction, and Usability section reports learners’ experiences with ChatGPT, including some students’ tendency to rely on its output without critically engaging with it. The cited paper frames ChatGPT through sociocultural mediation and scaffolding in its Theoretical Framework; it does not identify a fluency illusion or processing fluency as the mechanism behind those experiences. The review’s conceptualization is its own, but the source supports the narrower point about learners’ experiences with an external tool. (LLM, conf=0.83)

    Experimental — accuracy not yet measured

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentz = 0.17reported p = .881 · recomputed p = .865

    “,Z = .17,p = .881”

    The paper states an exact rank test

  • Consistentz = 0.17reported p = .862 · recomputed p = .865

    “,Z = .17,p = .862”

  • Consistentz = 0.47reported p = .639 · recomputed p = .638

    “,Z = .47,p = .639”

  • Consistentz = 2.64reported p = .008 · recomputed p = .008

    “,Z = 2.64, p = .008”

  • Consistentz = 2.59reported p = .01 · recomputed p = .010

    “,Z = 2.59, p = .010”

  • Consistentz = 1.13reported p = .257 · recomputed p = .258

    “,Z = 1.13, p = .257”

  • Consistentz = 2.16reported p = .031 · recomputed p = .031

    “(Z = 2.16, p = .031”

  • Consistentz = 4.86reported p = 0 · recomputed p < .001

    “(Z = 4.86, p = .000”

  • Consistentz = 4.89reported p = 0 · recomputed p < .001

    “(Z = 4.89, p = .000”

  • Consistentz = 2.23reported p = .026 · recomputed p = .026

    “,Z = 2.23, p = .026”

  • Consistentz = 2.21reported p = .027 · recomputed p = .027

    “,Z = 2.21, p = .027”

  • Consistentz = 1.47reported p = .141 · recomputed p = .142

    “,Z = 1.47, p = .141”

12 of 12

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Individual and intact-class random assignment conflict

    Minor

    The participants section says the 71 learners were randomly assigned to groups, while the procedure says two intact classes were randomly assigned. These describe different units of randomization.

  • Baseline overall-score Z differs between text and appendix

    Minor

    The main results report Z = .17 for baseline overall vocabulary scores, but Appendix D reports .15 for the same U statistic and p-value.

  • Delayed-test table labels its scores as pretest

    Minor

    Table 3 is identified as an analysis of delayed-test differences, but its score column is labeled "Pretest."

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 5

  1. Allen, D. (2004). Oxford Placement Test. Oxford University Press.
  2. v12n5p55 · DOI · Paper page
  3. Chen, X., Zou, D., Xie, H., Cheng, G., & Liu, C. (2022). Two decades of artificial intelligence in Eeducation: Contributors, collaborations, research topics, challenges, and future directions. Educational Technology & Society, 25(1), 28-47. https://www.jstor.org/stable/48647028
  4. Tilbury, A., Clementson, T., Hendra, L., & Rea, D. (2017). English unlimited: Special edition 3. Cambridge University Press.
  5. Vygotsky, L. S. (1978) Mind in society: The development of higher mental processes. Harvard University Press.

Cited by

Sorted by citations

Related

Authors
Abdelhalim, S. M., & Alsehibany, R.
Affiliations
Imam Mohammad ibn Saud Islamic University (SA)
ORCID
Venue
Language Learning & Technology
Publisher
National Foreign Language Resource Center
Published
2025-07-07
Pages
1-24
ISSN
1094-3501 (ISSN-L)
DOI
10.64152/10125/73635
OpenAlex
W4411988811
Semantic Scholar
37e55dd383a90988d5a10099f07d7bfcbe89b1fe
Cited by
17
References
55
Citations by year
2026: 13 · 2025: 4
FWCI
32.92
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Language learning & technology (publishedVersion, cc-by-nc-nd)
  • ScholarSpace (University of Hawaii at Manoa) (submittedVersion, cc-by-nc-nd)
Funders
Al-Imam Muhammad Ibn Saud Islamic University
Topics
  • Social Sciences › Psychology › Developmental and Educational Psychology › Second Language Acquisition and Learning
  • Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
  • Physical Sciences › Computer Science › Information Systems › English Language Learning and Teaching
Keywords
ChatGPT, vocabulary acquisition, vocabulary retention, EFL learners, Saudi students, AI-assisted language learning, expressive vocabulary, receptive vocabulary, Vygotsky's sociocultural theory, teacher-centered instruction, student motivation, student engagement, delayed recall

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-23.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-23.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review