Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth
Open accessThe DOI matches its Crossref record; 82 of the 84 references it cites were found in a registry or an index; 121 of the 137 sources its claims cite were checked for support.
Abstract
This systematic review synthesizes empirical evidence from 19 studies examining AI (AI) integration in academic writing assessment across higher education contexts. Following PRISMA 2020 guidelines, the review addresses five interconnected research questions regarding AI tool functionality, stakeholder perceptions, student motivations, institutional implementation gaps, and assessment redesign strategies. Findings reveal that generative AI platforms, particularly ChatGPT (88.8% adoption) and Grammarly (67.4%), dominate across writing stages, with students strategically deploying tools based on cognitive complexity levels. Stakeholder perspectives diverge significantly: students value productivity enhancement and accessibility, faculty prioritize academic integrity concerns, and administrators adopt cautious openness with disproportionate resource allocation favoring faculty guidance (67%) over direct student support (17.8%). Student motivations for inappropriate use stem from structural vulnerabilities including time pressures, ethical ambiguity, and institutional policy voids rather than malicious intent. Geographic inequities persist, with Global South institutions facing infrastructure limitations that amplify educational disparities. Promising pedagogical and assessment innovations include process-oriented assessment, oral defenses, multimodal composing, and explicit AI literacy curricula—each targeting the redesign of writing evaluation practices to balance AI assistance with authentic skill development. The review concludes that effective integration requires moving beyond detection-focused approaches toward comprehensive pedagogical redesign grounded in equity, authentic learning, and development of uniquely human capacities for critical judgment and ethical reasoning.
Overview
CoChat Summary
TL;DR
This PRISMA 2020 systematic review synthesizes 19 empirical studies on AI integration in higher-education academic writing assessment. It finds stratified AI tool use, divergent stakeholder views, structural drivers of inappropriate use, geographic inequities, and assessment redesign strategies intended to balance AI support with authentic skill growth.
Key findings
- ChatGPT dominated reported use at 88.8%, followed by Grammarly at 67.4%; Canva (57.3%) and Quizlet (30.3%) were reported in specific contexts. Tools were grouped into generative AI, automated writing evaluation, translation, and multimodal/interactive categories, with LLMs used more for higher-order planning/drafting and AWE tools for revision/editing.
- Stakeholder perspectives diverged: students valued productivity and accessibility, faculty prioritized integrity concerns, and administrators adopted “open but cautious” approaches. A resource gap was reported: 67% of guidance targeted faculty education while only 17.8% provided direct student frameworks.
- Inappropriate AI use was attributed less to malicious intent than to structural vulnerabilities including time pressures, ethical ambiguity, and institutional policy voids; “metacognitive laziness” emerged when assessment emphasized final products over learning processes.
- Geographic inequities persisted, with Global South institutions facing infrastructure limitations and research concentrated in Global North contexts. Promising redesign strategies included oral defenses, process-oriented assessment, multimodal composing, project-based learning, scaffolded assignments, and explicit AI literacy curricula.
Methods
- PRISMA 2020 systematic review searching Web of Science, Scopus, ERIC, PsycINFO, and Google Scholar for peer-reviewed English literature published January 2020–December 2025, using three thematic search clusters and forward/backward citation tracking.
- Inclusion required empirical research on AI in higher-education academic writing assessment, including student, faculty, or institutional perspectives; exclusions covered K-12-only studies, non-writing AI applications, purely theoretical pieces, dissertations/theses/preprints, non-English publications, and pre-2020 automated essay scoring without generative AI context.
- Screening began with 1,756 database records plus 18 citation records; after removing 385 duplicates, 1,389 records were screened, 1,251 were excluded, 138 full texts were assessed, 119 were excluded, and 19 studies were included. Inter-rater reliability was Cohen’s κ = 0.87; heterogeneity led to narrative synthesis, with quality appraisal using adapted Newcastle–Ottawa Scale, CASP, and MMAT.
Limitations
- The paper states that the 19-study corpus is small; evidence relies heavily on convenience sampling and self-reported data; longitudinal designs are absent; geographic concentration limits transferability; rapid AI evolution creates temporal instability; and publication bias favors positive-utility findings, with possible time-lag bias and marginalization of student voices.
- Evident from the text: the review uses narrative synthesis rather than meta-analysis because of heterogeneous designs, contexts, populations, and outcomes, and quality ratings were applied descriptively rather than to exclude studies, so conclusions rest on a methodologically varied evidence base.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
Verification
Claims the paper makes
68 Valid · 9 Not valid · 5 Not checked · 23 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Geographic diversity across Global North and South contexts strengthens external validity (Hanh and Duyen, 2025; Jeilani and Abubakar, 2025; Khlaif et al., 2025).”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Results > Publication and methodological bias assessment
What the sources say
- Disputedcites Hanh and Duyen, 2025 sourceExperimental — accuracy not yet measured
Section 3.6 (Limitations; page unavailable) says the study drew on a single postgraduate cohort, limiting diversity, and that its findings should not be treated as generalizable. Section 3.2 (Research design; page unavailable) likewise says its aim was not to generate generalizable claims. This paper therefore does not provide the claimed geographic diversity or resulting external validity, though the other cited studies are not assessed here. (LLM, conf=0.84)
Checked by a reasoning model — may be wrong.
- Partly supportedcites Jeilani and Abubakar, 2025 sourceExperimental — accuracy not yet measured
The Limitations section (section 8; page unavailable) says the study’s focus on Somalia may limit generalizability and calls for cross-cultural, multi-regional studies. That supports the value of geographic breadth, but the paper does not specifically discuss Global North–South diversity or show that such diversity strengthened external validity. (LLM, conf=0.86)
- Could not verifycites Khlaif et al., 2025 sourceExperimental — accuracy not yet measured
Section 3.7 (Trustworthiness; page unavailable) links purposive sampling to transferability across similar contexts, but does not link geographic diversity to external validity. Searches for Global North/South contexts, geographic locations, and generalizability found no passage making the attributed point. (LLM, conf=0.75)
boostedstates 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.
- Disputed
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Multi-source assessment ecosystems combining self-assessment, peer evaluation, AI analytics, and instructor judgment provide triangulated evidence of competence (Elshall and Badir, 2025; Francis et al., 2025; Tang et al., 2024; Zhang and Zhang, 2022).”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Introduction
What the sources say
- Partly supportedcites Elshall and Badir, 2025 sourceExperimental — accuracy not yet measured
Section 2.1 describes FACT as a multicomponent framework for comprehensive assessment, and section 4.5.2 says instructors can use its AI-free and AI-assisted tasks to monitor students’ independent skills. The paper does not describe an assessment ecosystem combining self-assessment, peer evaluation, AI analytics, and instructor judgment, or claim that those sources provide triangulated evidence of competence. (LLM, conf=0.85)
- Partly supportedcites Francis et al., 2025 sourceExperimental — accuracy not yet measured
The Competency-Based Learning section describes assessing observable competencies and using GenAI to provide performance feedback. Re-Thinking Assessment Methods also discusses instructor feedback and GenAI acting as an examiner. Neither section describes combining self-assessment, peer evaluation, AI analytics, and instructor judgment to produce triangulated evidence of competence; page numbers are unavailable in our copy. (LLM, conf=0.82)
- Partly supportedcites Tang et al., 2024 sourceExperimental — accuracy not yet measured
Section 2.1 recommends integrating feedback from learners, peers, teachers, and AI into writing analytics tools, and the Abstract identifies the lack of such integration as a design gap. The paper does not establish that an ecosystem combining all four provides triangulated evidence of competence; its discussion concerns formative writing feedback rather than competence assessment generally. No page numbers were available in our copy. (LLM, conf=0.84)
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.
- Rests on an assumption not shown here.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Positioning AI as feedback partner rather than ghostwriter, with structured requirements to critique and adapt AI-generated suggestions, enhances metacognitive engagement and critical evaluation capacities (Akiba and Garte, 2024; Khojasteh et al., 2025; Oates and Johnson, 2025; Song and Song, 2023).”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Introduction
What the sources say
- Partly supportedcites Akiba and Garte, 2024 sourceExperimental — accuracy not yet measured
The Assignment Structure and Procedure section requires students to revise drafts using AI feedback and explain why they accepted or rejected suggestions. The Discussion reports active engagement beyond surface edits and describes AI as a writing partner rather than a shortcut. The study does not establish that this approach enhances metacognitive engagement or critical evaluation capacities as distinct outcomes. (LLM, conf=0.85)
- Supportedcites Khojasteh et al., 2025 source
In Results and discussion (page not supplied), the paper says its self-assessment prompts guide students to engage critically with AI feedback and nurture metacognitive skills. Its practical guidelines recommend using AI feedback as a supplement, not a replacement, and asking students to explain which suggestions they accept or reject. (LLM, conf=0.90)
- Partly supportedcites Oates and Johnson, 2025 sourceExperimental — accuracy not yet measured
The Introduction describes a structured task in which students generated an essay with ChatGPT and critically evaluated it using peer-reviewed sources; the Discussion says critiquing AI output may foster critical evaluation skills rather than replace human effort. The paper does not establish that this approach enhances metacognitive engagement, and direct AI feedback to help students refine their evaluations is proposed as a further stage, not reported as a tested result. Page numbers are unavailable in our copy. (LLM, conf=0.84)
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.
- Rests on an assumption not shown here.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Authentic, cognitively demanding tasks incorporating oral defenses, experiential components, and disciplinary problem-solving help preserve academic integrity while building employability skills (Salinas-Navarro et al., 2024; Sotiriadou et al., 2019; Villarroel et al., 2019).”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Introduction
What the sources say
- Partly supportedcites Salinas-Navarro et al., 2024 sourceExperimental — accuracy not yet measured
The Introduction (p. 2) links experiential, higher-order assessments to concerns about plagiarism and says such tasks are harder for GenAI tools to handle. Sections 2.1–2.2 (pp. 3–4) connect experiential learning and disciplinary problem-solving with capabilities relevant to future professional careers. The paper does not identify oral defenses as a component of this approach, so it does not support that specific part of the claim. (LLM, conf=0.85)
- Supportedcites Sotiriadou et al., 2019 source
The Abstract reports that scaffolded authentic assessments culminating in interactive oral exams helped prevent academic misconduct and develop skills relevant to employability. The Phase 1: Sport Management Course section describes oral questions requiring students to solve discipline-specific organisational problems, while the International Management Course section describes the assessments as experiential. (LLM, conf=0.88)
- Not checked — no full text of the sourcecites Villarroel et al., 2019 source
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.
- Rests on an assumption not shown here.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Memory retention and autonomous problem-solving face particular risks when students adopt AI as answer-provider rather than thinking partner (Gerlich, 2025; Krsmanović and Deek, 2025; Zhai et al., 2024).”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Introduction
What the sources say
- Supportedcites Gerlich, 2025 source
Section 2.4 (page not specified) warns that relying on AI to offload cognitive tasks could erode memory retention and independent problem-solving. Sections 2.5 and 5.2 (pages not specified) apply the concern to students, contrasting active engagement with passive reliance on AI tools. (LLM, conf=0.88)
- Supportedcites Krsmanović and Deek, 2025 source
The Discussion warns that using AI to answer assignments instead of seeking deeper understanding can let students avoid thinking through the material. It also reports self-perceived memorization problems and cautions that AI must not replace students’ cognitive and problem-solving skills. The claim describes risks, not a demonstrated causal comparison between ways of using AI. (LLM, conf=0.90)
- Partly supportedcites Zhai et al., 2024 sourceExperimental — accuracy not yet measured
The Introduction and Discussion (pages unavailable) warn that relying on AI-generated answers may discourage students from forming their own insights and weaken independent problem-solving. The paper contrasts this with using AI to stimulate critical thinking. It does not establish a particular risk to students’ memory retention; its discussion of memory formation concerns the hippocampus’s role in decision-making, not an effect of AI use. (LLM, conf=0.86)
plainstates a possible 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.
- Rests on an assumption not shown here.
- Supported
54 more claims · 42 with their cited sources checked
Create a free account to see all 105 claims and to verify the 2 references still open.
- It may state two separate claims as one. 34
- Probably not a substantial claim. 17
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 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
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Materials and methods > Search strategy
“This five-year window was strategically selected”
§ Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth > Materials and methods > Search strategy
“published between January 2020 and December 2025.”
Citations
Cites
The PRISMA 2020 statement: An updated guideline for reporting systematic reviews
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., et al. (2021) · BMJ · doi:10.1136/bmj.n71 · Cited by 108,283
Cognitive Load During Problem Solving: Effects on Learning
Sweller, J. (1988) · Cognitive Science · doi:10.1016/0364-0213(88)90023-7 · Cited by 11,048
A revision of bloom's taxonomy: an overview
Krathwohl, D. R. (2002) · Theory Into Practice · doi:10.1207/s15430421tip4104_2 · Cited by 8,132 · cited 1×
“…logic provides the conceptual entry point for Bloom’s Taxonomy (Anderson and Krathwohl, 2001), which organizes cognitive work along a hierarchy from lower-order skills—remember-ing, understanding, and applying—to higher-order skills such as analyzing, evaluating, and creating (Krathwohl, 2002).”
Cognitive Load Theory and Instructional Design: Recent Developments
Paas, F., Renkl, A., & Sweller, J. (2003) · Educational Psychologist · doi:10.1207/s15326985ep3801_1 · Cited by 3,041
Students’ voices on generative AI: perceptions, benefits, and challenges in higher education
Chan, C. K. Y., & Hu, W. (2023) · International Journal of Educational Technology in Higher Education · doi:10.1186/s41239-023-00411-8 · Cited by 2,301 · cited 7×
“…the primary end-users, directly experiencing AI’s impact on their learning processes, academic integrity concerns, and sense of authorship (Chan and Hu, 2023; Kim et al., 2024; Lin and Chen, 2024; Song and Song, 2023; Wang, 2024), with multilingual learners particularly affected by…”
InfluentialArtificial intelligence in higher education: the state of the field
Crompton, H., & Burke, D. (2023) · International Journal of Educational Technology in Higher Education · doi:10.1186/s41239-023-00392-8 · Cited by 2,010 · cited 3×
“Faculty members occupy a pivotal decision-making role, determining acceptable AI use in assignments while balancing pedagogical benefits against integrity and creativity concerns (Crompton and Burke, 2023; Gustilo et al., 2024; Khlaif et al., 2024).”
A comprehensive AI policy education framework for university teaching and learning
Chan, C. K. Y. (2023) · International Journal of Educational Technology in Higher Education · doi:10.1186/s41239-023-00408-3 · Cited by 1,577 · cited 4×
“These inequities stem from limited digital infrastructure, policy gaps, English-centric tool design, and socio-cultural barriers (Chan, 2023; Gustilo et al., 2024; Kalim et al., 2025).”
InfluentialZhai, C., Wibowo, S., & Li, L. D. (2024) · Smart Learning Environments · doi:10.1186/s40561-024-00316-7 · Cited by 1,567 · cited 5×
“Memory retention and autonomous problem-solving face particular risks when students adopt AI as answer-provider rather than thinking partner (Gerlich, 2025; Krsmanović and Deek, 2025; Zhai et al., 2024).”
InfluentialAI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking
Gerlich, M. (2025) · Societies · doi:10.3390/soc15010006 · Cited by 1,248 · cited 3×
“…and basic application (Abdelmagid et al., 2025; Essien et al., 2024; Zhou et al., 2024), while higher-order skills including critical thinking, analysis, evaluation, and creative synthesis prove more vulnerable to erosion through cognitive offloading (Gerlich, 2025; Zhai et al., 2024).”
AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking
Gerlich, M. (2025) · doi:10.2139/ssrn.5082524 · Cited by 1,238
Research works printed but not linked: 2
- Anderson, 2001, A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Educational Objectives
- Talandis, 2025, Moving JALT into the Future: Opportunity, Diversity, and Excellence
Metadata
- Authors
- Abdellatif, M. S., Alshehri, M. A., Lamouchi, A., Rahmath, M., & Nemt-allah, M. A.
- Affiliations
- Prince Sattam Bin Abdulaziz University (SA)
- King Saud University (SA)
- Al-Azhar University (EG)
- Corresponding author
- Nemt-allah, Mohamed Ali
- ORCID
- Venue
- Frontiers in Psychology
- Publisher
- Frontiers Media SA
- Published
- 2026-05-26
- Volume
- 17
- Pages
- 1809174-1809174
- ISSN
- 1664-1078 (ISSN-L)
- PMC
- PMC13246662
- OpenAlex
- W7162420461
- Semantic Scholar
- 5a16a42a0b9e91f2b0e84b85c42b1f03a3e840ed
- Cited by
- 0
- References
- 80
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- 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)
- Topics
- Social Sciences › Education › Artificial Intelligence in Education
- Social Sciences › Education › Academic integrity and plagiarism
- Social Sciences › Education › Student Assessment and Feedback
- Keywords
- academic writing, generative AI tools, ChatGPT, Grammarly, AI literacy, academic integrity, student motivation, stakeholder perspectives, assessment redesign, process-oriented assessment, multimodal composition, AI detection, educational equity, 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).
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: Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth. https://app.cochat.ai/papers/10.3389/fpsyg.2026.1809174
- The paper
- Abdellatif, M. S., Alshehri, M. A., Lamouchi, A., Rahmath, M., & Nemt-allah, M. A. (2026). Reimagining writing assessment for the AI era: a systematic review on balancing AI support and authentic skill growth. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1809174
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.
