“Scarlet Cloak and the Forest Adventure”: a preliminary study of the impact of AI on commonly used writing tools
Open accessThe DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; 2 of the 13 sources its claims cite were checked for support.
Abstract
The findings reveal that tools like Grammarly that subtly enhance readability also trigger detection and increase false positives, especially for non-native speakers, and recommends a shift toward teaching responsible AI usage rather than pursuing rigid bans or relying on detection metrics that may not accurately capture misconduct.
Overview
CoChat Summary
TL;DR
This preliminary study applies AI writing and paraphrasing tools to three pre-LLM texts and finds that common aids like Grammarly can trigger false positives in Turnitin and GPTZero, while some paraphrasing tools and Microsoft Bing Copilot often evade detection. It argues that AI detectors cannot reliably distinguish legitimate paraphrasing from AI generation and recommends managing AI use through policy and transparency rather than bans.
Key findings
- Grammarly modifications triggered detection: on Perrault’s “Little Red Riding Hood,” Turnitin scored 11% and GPTZero 33%, versus 0% and 0% for the original; on Barbrook et al., GPTZero scored 32%, while O’Donnell scored 0% on GPTZero. The abstract states such false positives especially affect non-native speakers.
- GPT 3.5 and 4 were often flagged by GPTZero: GPT 3.5 scored 65% on O’Donnell, 48% on Barbrook et al., and 86% on Perrault; Turnitin remained mostly 0%, except 11% on O’Donnell. GPT 4 on Perrault scored 48% on GPTZero and 0% on Turnitin.
- Writefull and Microsoft Bing Copilot often escaped detection: Writefull had 0% Turnitin across texts and GPTZero scores of 2%, 11%, and 1%; Copilot on Perrault had 0% Turnitin and 0% GPTZero. EditPad generally scored low on GPTZero (0–33%) but produced extreme perplexity and burstiness, including 4701.4 burstiness on Perrault.
- Turnitin and GPTZero could not definitively detect AI interventions: Turnitin hardly revealed AI interventions in most cases, while GPTZero remained imperfect, especially for paraphrased content.
Methods
- The study used three texts written before consumer LLM tools: Barbrook et al. (1998, *Nature*), O’Donnell (2009, *Heroic Age*), and Perrault’s “Little Red Riding Hood” (1891). Texts were converted to plain .txt files, with bibliographies and meta-textual elements removed.
- Tools applied were ChatGPT GPT-3.5 and GPT-4, Microsoft Bing Copilot, EditPad, Grammarly, and Writefull, focusing on paraphrasing. Each tool required different input procedures and prompts; outputs were stored in a GitHub repository.
- Original and rewritten texts were run through Turnitin and GPTZero. The authors compared detection scores and metrics including readability, SAT percentage, simplicity, perplexity, burstiness, and average sentence length, producing heat maps, scatter plots, histograms, and area graphs.
Limitations
- The paper states the dataset comprises only three texts, limiting broader conclusions; possible self-selection bias affects generalizability; the novel exploratory methodology has not been validated in prior research; and genre/style factors require future research with a larger, more diverse corpus.
- The paper also states that current AI detection tools predominantly focus on English-language outputs, leaving a gap in understanding performance across other languages.
- Evident from the text: results are tied to specific tool versions, prompts, and chunking constraints (e.g., Copilot’s 4000-character limit, Writefull’s 40–2000-character inputs, EditPad’s instability), and no inferential statistics are reported, so findings may not generalize.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 6 Not valid · 8 Not checked · 94 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Current AI detection metrics still struggle to reliably assess texts that have been partially processed or enhanced by tools like ProWritingAid, as the resulting quality differs subtly from that generated entirely by mainstream LLMs like ChatGPT or Llama 2 (Elkhatat et al., 2023).”
What the sources say
- Partly supportedcites Elkhatat et al., 2023 sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Additionally, they also have a problem with false positives (i.e. mistakenly identifying human-generated writing as AI-generated), particularly with writing produced by non-native speakers of English and members of Equity-Deserving Groups (Chechitelli, 2023; Quidwai et al., 2023).”
What the sources say
- Not matchedcites Chechitelli, 2023
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Partly supportedcites Quidwai et al., 2023 sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Turnitin gave our abstract a score of 16% for AI generation”
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In general, paraphrasing tools score low values in AI detection software, allowing the changes to go mostly unnoticed by the software.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The results highlight that widely used writing aids, even those not primarily generative, can trigger false positives in AI detection tools.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
53 more claims
Create a free account to see all 110 claims and to verify the 2 references still open.
- Probably not a substantial claim. 23
- It may state two separate claims as one. 39
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.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Discussion · p. 17
“In all of the cases analyzed, Writefull and Grammarly seem to avoid being picked up by either detector completely.”
§ Reviewing the results in Turnitin and GPTZero · p. 13
“| Writefull | 0 | 11 | 22.7 | 2.8 | 31.4 | 71.1 | 79.4 | 17.1 | | Grammarly | 0 | 32 | 21.1 | 5.1 | 36.7 | 65.8 | 76.1 | 22.4 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Editpad · p. 14
“17.7 in Grammarly-modified O'Donnell”
§ Reviewing the results in Turnitin and GPTZero · p. 12
“| Grammarly | 0 | 0 | 33.0 | 3.7 | 40.2 | 70.8 | 373.1 | 16.5 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Grammarly · p. 14
“65.8 perplexity and 76.1 burstiness in Perrault”
§ Reviewing the results in Turnitin and GPTZero · p. 13
“| Gram- marly | 11 | 33 | 70.2 | 0.8 | 39.4 | 48.0 | 103.3 | 18.3 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Writefull · p. 15
“71.1 and 85.0 in Perrault”
§ Reviewing the results in Turnitin and GPTZero · p. 13
“| Writefull | 0 | 1 | 74.2 | 0.5 | 38.7 | 55.7 | 85.0 | 16.3 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Editpad · p. 14
“Its simplicity scores vary from 34.4 to 39.5”
§ Reviewing the results in Turnitin and GPTZero · p. 12
“| EditPad | 0 | 2 | 17.7 | 3.6 | 41.6 | 190.2 | 321.6 | 30.9 |”
3 more findings
Create a free account to see every finding of this check.
Citations
Cites
BAİDOO-ANU, D., & OWUSU ANSAH, L. (2023) · Journal of AI · doi:10.61969/jai.1337500 · Cited by 3,939
Baidoo-Anu, D., & Owusu Ansah, L. (2023) · doi:10.2139/ssrn.4337484 · Cited by 3,004
Ray, P. P. (2023) · Internet of Things and Cyber-Physical Systems · doi:10.1016/j.iotcps.2023.04.003 · Cited by 2,463 · cited 1×
“While ChatGPT can pass some subjects and excel in certain assessment types, current assessment practices may need to change and require significant adaptation to withstand these advancements (Ray, 2023; Nikolic et al., 2023).”
Testing of detection tools for AI-generated text
Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Šigut, P., & Waddington, L. (2023) · International Journal for Educational Integrity · doi:10.1007/s40979-023-00146-z · Cited by 582 · cited 1×
“While these AI-powered writing tools offer great benefits, they have also introduced new challenges in academic settings, particularly in the assessment of student work (Adams & Chuah, 2022; Baidoo-Anu & Ansah, 2023; Debora et al., 2023; Giray, 2024; Lambert & Stevens, 2023; Perkins et al., 2023).”
Human-AI collaboration patterns in AI-assisted academic writing
Nguyen, A., Hong, Y., Dang, B., & Huang, X. (2024) · Studies in Higher Education · doi:10.1080/03075079.2024.2323593 · Cited by 449 · cited 1×
“…of various applications, including writing assistants, i.e., software programs that help writers with tasks such as spelling correction, grammar checking, readability analysis, feedback generation, and/or even plagiarism detection (Giray, 2024; Nguyen et al., 2024; O’Neill & Russell, 2019).”
Elkhatat, A. M., Elsaid, K., & Almeer, S. (2023) · International Journal for Educational Integrity · doi:10.1007/s40979-023-00140-5 · Cited by 392 · cited 1×
“Current AI detection metrics still struggle to reliably assess texts that have been partially processed or enhanced by tools like ProWritingAid, as the resulting quality differs subtly from that generated entirely by mainstream LLMs like ChatGPT or Llama 2 (Elkhatat et al., 2023).”
Nikolic, S., Daniel, S., Haque, R., Belkina, M., Hassan, G. M., Grundy, S., Lyden, S., Neal, P., & Sandison, C. (2023) · European Journal of Engineering Education · doi:10.1080/03043797.2023.2213169 · Cited by 336 · cited 1×
“While ChatGPT can pass some subjects and excel in certain assessment types, current assessment practices may need to change and require significant adaptation to withstand these advancements (Ray, 2023; Nikolic et al., 2023).”
Perkins, M., Roe, J., Postma, D., McGaughran, J., & Hickerson, D. (2024) · Journal of Academic Ethics · doi:10.1007/s10805-023-09492-6 · Cited by 234
Game of Tones: Faculty detection of GPT-4 generated content in university assessments
Perkins, M., Roe, J., Postma, D., McGaughran, J., & Hickerson, D. (2023) · arXiv · doi:10.48550/arxiv.2305.18081 · Cited by 233
The phylogeny of The Canterbury Tales
Barbrook, A. C., Howe, C. J., Blake, N. F., & Robinson, P. M. W. (1998) · Nature · doi:10.1038/29667 · Cited by 150 · cited 2×
“A multi-author interdisciplinary article focusing on a humanities topic but published in the primarily science-focused journal Nature that was written in part by a close collaborator of our lab (Barbrook et al., 1998); 2.”
Research works printed but not linked: 2
- Kuykendall, K. (2023, May 30). Turnitin AI detector analyzed 38M submissions in its first 6 weeks; updates answer educator feedback. Technological Horizons in Education Journal (THE). https://thejournal.com/Articles/2023/05/30/Turnitin-Shares-Stats-From-AI-Detection-of-38M-Submissions-Tweaks-Detector-Feature.aspx
- O’Neill, R., & Russell, A. M. T. (2019). Grammarly: Help or hindrance? Academic learning advisors’ perceptions of an online grammar checker. Journal of Academic Language and Learning, 13(1), A88–A107.
Cited by
The Transparency Paradox: Why Researchers Avoid Disclosing AI Assistance in Scientific Writing
BaHammam, A. S. (2025) · Nature and Science of Sleep · doi:10.2147/nss.s568375 · Cited by 39
“16,17 The reluctance to disclose AI use becomes more understandable when considering the current limitations of detection technologies.”
Luis, S. Y., Reina, D. G., & Toral, S. L. (2025) · Education Sciences · doi:10.3390/educsci15060706 · Cited by 10
AI Detection in Academia: How Indian Universities Can Safeguard Academic Integrity
Gupta, A., Mahaseth, H., & Bajpai, A. (2025) · International Global Conference Series on ICT Integration in Technical Education & Smart Society · doi:10.3390/engproc2025107026 · Cited by 9
AI writing detectors are ineffective, unreliable and harmful
Giray, L., Roe, J., & Espiritu, J. D. (2026) · English Teaching Practice & Critique · doi:10.1108/etpc-07-2025-0155 · Cited by 8
Anthony, A., & Sharma, S. (2025) · Journal of Educational Technology Systems · doi:10.1177/00472395251356520 · Cited by 5
Metadata
- Authors
- Bordalejo, B., Pafumi, D., Onuh, F., Khalid, A. K. M. I., Pearce, M. S., & O’Donnell, D. P.
- Affiliations
- University of Lethbridge (CA)
- Corresponding author
- Pafumi, Davide
- Venue
- International Journal of Educational Technology in Higher Education
- Published
- 2025-02-07
- Volume
- 22
- Issue
- 1
- ISSN
- 2365-9440 (ISSN-L)
- OpenAlex
- W4407203081
- Cited by
- 13
- References
- 24
- Citations by year
- 2026: 8 · 2025: 5
- FWCI
- 13.59
- Type
- journal article
- Language
- en
- Update notices
- correction (2025-03-12)
- Open access
- diamond
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- International Journal of Educational Technology in Higher Education (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Social Sciences › Education › Academic integrity and plagiarism
- Social Sciences › Education › Artificial Intelligence in Education
- Physical Sciences › Computer Science › Artificial Intelligence › Authorship Attribution and Profiling
- Keywords
- generative AI detection, writing tools, Grammarly, Turnitin, GPTZero, academic integrity, readability, perplexity, burstiness, non-native speakers, responsible AI use, AI policy in higher education
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 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
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: “Scarlet Cloak and the Forest Adventure”: a preliminary study of the impact of AI on commonly used writing tools. https://app.cochat.ai/papers/10.1186/s41239-025-00505-5
- The paper
- Bordalejo, B., Pafumi, D., Onuh, F., Khalid, A. K. M. I., Pearce, M. S., & O’Donnell, D. P. (2025). “Scarlet Cloak and the Forest Adventure”: a preliminary study of the impact of AI on commonly used writing tools. International Journal of Educational Technology in Higher Education. https://doi.org/10.1186/s41239-025-00505-5
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.