Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 32 of the 34 references it cites were found in a registry or an index; 27 of the 65 sources its claims cite were checked for support.

Perceptions vs. practices: academic integrity and actual ChatGPT use among EFL students

Open access
Alsadoon, Reem · 2026 · Frontiers in Psychology

Abstract

This mixed-methods study explores how English as a Foreign Language (EFL) students and their teachers think about and actually use ChatGPT in academic writing, and how this relates to academic integrity. A unique contribution of the study is its use of observed behavioral data, capturing students' real-time interactions with ChatGPT rather than relying solely on self-reports. 60 female Saudi EFL students wrote a timed argumentative essay in a computer lab where they were allowed to use ChatGPT. Screen recordings captured all their prompts and AI outputs. These texts were later compared with students' final essays using copy ratio and Levenshtein distance to show how much they copied or revised. Students also completed pre- and post-task surveys about their ChatGPT use and integrity beliefs, and 20 teachers rated how acceptable eight ChatGPT functions were for academic writing. Short interviews were conducted with a sample of students. Findings showed that students viewed ChatGPT more positively than teachers and spent about one third of the writing time using it, mainly for editing and brainstorming. Cluster analysis identified three user types: constructive, hybrid, and substitutive, with about one-fifth of students relying heavily on AI text. Higher copying was linked to weaker integrity beliefs, while more revision and clearer policy awareness were linked to stronger ethical views. The study calls for clearer AI policies, process-based assessment, and practical AI literacy training.

Overview

CoChat Summary

TL;DR

This mixed-methods lab study observed 60 female Saudi EFL students using ChatGPT during a timed essay and compared their behavior with student and teacher integrity perceptions. It found most used ChatGPT constructively, but about one-fifth were substitutive users, and higher copying was linked to weaker integrity beliefs while teachers were stricter than students about high-risk functions.

Key findings

  • Students spent M = 14.8 min (SD = 6.3) of the 45-min session using ChatGPT, about 33% of writing time, with 512 recorded prompts (8.5 per student); the most frequent functions were editing (22%), brainstorming (20%), and drafting (16% combined).
  • Cluster analysis identified constructive (42%), hybrid (38%), and substitutive (20%) users; overall copy ratio was M = 0.36 (SD = 0.18), with constructive users at CR ≈ 0.22, hybrid at ≈ 0.39, and substitutive at ≈ 0.64, F(2, 57) = 28.9, p < .001, η² = 0.50.
  • Copy ratio correlated negatively with integrity-belief scores (r = −.46, p < .01), while Levenshtein distance correlated positively (r = .42, p < .01); regression explained 39% of variance, and high-dependence students (CR ≥ 0.35) had lower integrity scores (M = 3.58) than low-dependence students (M = 4.06), t(58) = 4.17, p < .001, d = 0.91.
  • Students rated ChatGPT use as more acceptable than teachers (M = 3.87 vs. 3.25; t(78) = 5.13, p < .001, d = 1.15); teachers rated drafting, translation, and paraphrasing as significantly less acceptable, and the drafting acceptability gap correlated with copy ratio (r = .49, p < .001).

Methods

  • Mixed-methods lab-controlled design: 60 female EFL students in Level 4 Essay Writing II at Imam Mohammed Ibn Saud Islamic University, Riyadh, wrote a 45-min argumentative essay (300–350 words) with ChatGPT allowed; 20 female teachers rated acceptability, and 10 students were interviewed.
  • Behavioral data came from OBS Studio screen recordings and a logging extension capturing prompts, time, and copy-paste events; 512 prompts were coded into eight functions (Cohen’s κ = 0.81), and final essays were compared with ChatGPT outputs using copy ratio and Levenshtein distance in R (stringdist, quanteda).
  • Surveys included a pre-task questionnaire, a post-task questionnaire (25 closed items; IBI α = .88), and a teacher questionnaire (24 items; α = .91); analysis used descriptives, t-tests, ANOVA, Pearson correlations, multiple regression, and thematic analysis.

Limitations

  • Stated in the paper: the sample was restricted to female participants from a single Saudi institution, limiting generalizability; the controlled laboratory design may not fully reflect naturalistic AI use; and short-term writing tasks limit conclusions about long-term language development.
  • Evident from the text: the correlational and regression analyses cannot establish causal direction between copying, revision, policy awareness, and integrity beliefs.
  • Evident from the text: the teacher sample (20) and interview subsample (10) are small relative to the student sample, and all participants were female, limiting stakeholder and gender comparisons.

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

Integrity signals

Registry

Crossref · 10.3389/fpsyg.2026.1796737

Cites

32 of 34 references verified1 citation errors · 1 not checkable

Statistics

Figures

Figure 1 Comparison of mean Copy Ratios (CR) among the three user types.
Figure 1 Comparison of mean Copy Ratios (CR) among the three user types.CC BY · PubMed Central
Figure 2 Scatterplots showing correlations between behavioral measures and integrity beliefs.
Figure 2 Scatterplots showing correlations between behavioral measures and integrity beliefs.CC BY · PubMed Central
Figure 3 Mean acceptability ratings for eight ChatGPT functions by group.
Figure 3 Mean acceptability ratings for eight ChatGPT functions by group.CC BY · PubMed Central
Figure 4 Mean acceptability ratings of eight ChatGPT functions among students and teachers.
Figure 4 Mean acceptability ratings of eight ChatGPT functions among students and teachers.CC BY · PubMed Central
Figure 5 Comparison of students' and teachers' acceptability ratings across ChatGPT functions in EFL writing.
Figure 5 Comparison of students' and teachers' acceptability ratings across ChatGPT functions in EFL writing.CC BY · PubMed Central

Verification

Claims the paper makes

21 Valid · 5 Not valid · 23 Not checked · 87 Own results and interpretations

70 more claims · 3 with their cited sources checked

Create a free account to see all 136 claims and to verify the 3 references still open.

See all claims — free account
  • Probably not a substantial claim. 42
  • It may state two separate claims as one. 35

Create a free account to see these claims.

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.

  • ConsistentF(2, 57) = 28.9reported p < .001 · recomputed p < .001

    “F(2, 57) = 28.9, p < .001”

  • ConsistentF(4, 55) = 8.67reported p < .001 · recomputed p < .001

    “F(4, 55) = 8.67, p < .001”

  • Consistentt(58) = 4.17reported p < .001 · recomputed p < .001

    “t (58) = 4.17, p < .001”

  • Consistentt(78) = 5.13reported p < .001 · recomputed p < .001

    “t (78) = 5.13, p < .001”

  • ConsistentF(1, 78) = 21.6reported p < .001 · recomputed p < .001

    “F(1, 78) = 21.6, p < .001”

5 of 6 · the rest aren’t loaded

1 more recomputed test

Create a free account to see every test the Stats check recomputed.

See all tests — free account

Self-check

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

Self-check results appear here once an independent check has confirmed them.

Citations

Cites

Show all 30 references

Research works printed but not linked: 5

  1. Awan, 2023, Quality standards of qualitative research in applied linguistics: a conceptual review, VFAST Trans. Educ. Soc. Sci., 11, 68
  2. Baskara, 2023, “ChatGPT-assisted English language learning: Theoretical implications for global mobility and cross-cultural communication,”, 105
  3. Huang, 2011, “EdIt: a broad-coverage grammar checker using pattern grammar,”, 26
  4. Kivrak, 2024, Exploring EFL students' perceptions of using ChatGPT as a complementary tool in writing classes, LET Journal, 4, 170
  5. Vygotsky, 1978, Mind In Society: the Development of Higher Psychological Processes

Cited by

See all 8

Metadata

Authors
Alsadoon, R.
Affiliations
Imam Mohammad ibn Saud Islamic University (SA)
Corresponding author
Alsadoon, Reem
Venue
Frontiers in Psychology
Published
2026-03-19
Volume
17
Pages
1796737-1796737
ISSN
1664-1078 (ISSN-L)
OpenAlex
W7138927951
Cited by
8
References
30
Citations by year
2026: 8
FWCI
40.83
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Psychology (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Deanship of Scientific Research, Imam Mohammed Ibn Saud Islamic University
Topics
  • Social Sciences › Education › Academic integrity and plagiarism
  • Social Sciences › Education › Artificial Intelligence in Education
  • Social Sciences › Arts and Humanities › Language and Linguistics › EFL/ESL Teaching and Learning
Keywords
ChatGPT, academic integrity, EFL learners, AI-assisted writing, student perceptions of AI, teacher perceptions, AI-generated text, AI literacy, argumentative essay writing, Saudi EFL learners

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Perceptions vs. practices: academic integrity and actual ChatGPT use among EFL students. https://app.cochat.ai/papers/10.3389/fpsyg.2026.1796737
The paper
Alsadoon, R. (2026). Perceptions vs. practices: academic integrity and actual ChatGPT use among EFL students. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1796737

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.