xPeerd (AI review)

November 9, 2026

AI Review (xPeerd — DAReview)

Summary

This manuscript examines whether perceived administrative support for guidance and counselling is associated with students’ academic adjustment and intended retention at the University of Skills Training and Entrepreneurial Development (USTED), Ghana. The study uses a cross-sectional survey distributed through WhatsApp/Google Forms and ultimately centers its inferential analyses on 65 student respondents, although 161 total responses were collected across students, lecturers/tutors, and support staff. The paper’s core finding is that academic adjustment is more strongly associated with intended retention than perceived administrative support when both are entered jointly in regression models. The manuscript is commendably transparent in acknowledging measurement overlap, repeated item wording, low internal consistency for the administrative-support index, and the limits of cross-sectional inference. It also identifies practically relevant institutional barriers such as lack of private counselling offices, limited counselling time, and shortages of trained personnel.

However, I find substantial concerns about manuscript integrity, readability, measurement validity, sampling logic, and the coherence of the written argument. The paper shows pervasive unnatural phrasing, repeated discourse fillers, grammatical corruption, and citation anomalies that materially interfere with scholarly clarity. Methodologically, the convenience sample, small analytic student subsample, mixed respondent frame, weak reliability of the key predictor, and overlap between predictor and outcome wording all limit interpretability. On balance, the study’s practical topic is relevant, but the manuscript requires major revision before it can support a strong empirical contribution.

Potential Major Revisions

1. Severe language-quality and coherence problems threaten interpretability and may indicate manuscript-generation artifacts

The manuscript contains persistent non-idiomatic, semantically awkward, and mechanically repetitive phrasing that undermines confidence in editorial control and, at points, in the precision of the authors’ claims. Examples include “this study inspected this relationship” in the Abstract , “intention to continue papering?” in the Introduction , “students who spotted as students and university staff” , and “the put forward paper therefore treats all all these five questions” in the literature section . Similar distortions recur throughout the paper: “veering levels of internal consistency” , “second reasonable level of administrative support” , and “the autonomous coefficient for administrative support” in the regression results .

This is not a minor copyediting issue. The density of corrupted syntax, filler expressions, and malformed academic prose makes it difficult to determine which formulations are intentional theoretical nuances and which are accidental textual noise. The manuscript should be comprehensively rewritten for grammatical correctness, semantic clarity, and discipline-appropriate style before scientific evaluation can fully proceed.

2. The sampling frame and analytic sample are not well aligned with the stated research objective

The paper’s substantive questions concern student academic adjustment and intended retention, yet the questionnaire was distributed to a broader pool including “students and university staff” and produced “161 responses,” of which only “65 student responses” were used as “the primary basis for student-outcome associations” . The Method reiterates that the sample included “students (65; 40.4%), lecturers/tutors (58; 36.0%) and supporting team (38; 23.6%)” .

This design raises several concerns. First, the rationale for surveying non-students is underdeveloped if the principal outcomes are inherently student self-reports. Second, the inclusion of staff in the survey introduction may have affected recruitment, item interpretation, and response context. Third, the paper does not explain whether duplicate access, eligibility verification, or institutional affiliation checks were implemented in the Google Forms/WhatsApp distribution process. Fourth, convenience sampling through WhatsApp groups creates strong selection risks that likely favor more digitally connected or institutionally engaged respondents. The limitations section acknowledges restricted generalisation due to “convenience sampling through WhatsApp groups” and the small student sample , but the manuscript should do much more to justify the design and delimit the claims accordingly.

3. The key administrative-support measure is conceptually heterogeneous and psychometrically weak

The paper explicitly states that guidance and counselling administrative support (GCAS) combines private meeting space, time allocation, resources, leadership support, and coordinator training into one averaged index, while also conceding that these are distinct administrative conditions rather than a homogeneous latent trait . The reported reliability is low: “Cronbach's α = .55” .

I appreciate the manuscript’s transparency that GCAS is “treated as a descriptive administrative-support index, not a validated one-factor latent measure” . Still, averaging heterogeneous components into a single predictor and then testing it in regression encourages overinterpretation of a composite that the paper itself admits may not behave as a coherent scale. If the construct is formative or descriptive rather than reflective, the analytic logic should change. The authors should consider item-level analyses, theoretically grouped subdomains, or a clearer justification for index construction. As written, the low alpha is not merely a caveat; it directly weakens the meaning of the reported coefficient estimates.

4. Measurement overlap between predictor and outcomes risks inflated associations

One of the manuscript’s most serious internal concerns is acknowledged by the authors themselves: “Some adjustment and retention statements unambiguously are mentioning school support or guidance as the reason for a favourable experience; this is creating an overlap between the predictor and the suggested outcomes” . The paper further notes that some repeated statements were removed from the primary composites and that exploratory shorter scores were created excluding support-attributing items .

This is important because the central argument depends on distinguishing administrative support from academic adjustment and intended retention. If some outcome items directly attribute positive adjustment or retention to support, common-method overlap may partly manufacture the observed relationships. The sensitivity analyses partially address this; for example, the “support-attribution items excluded” model still shows adjustment as significant while GCAS remains non-significant in the joint model . That helps, but the revised scores are themselves described as not “a verified scale” . The manuscript should present the exact questionnaire items in an appendix and more rigorously defend the final scoring decisions. Without that transparency, the construct boundaries remain uncertain.

5. Cross-sectional regression is discussed carefully in places, but causal language still requires tighter control

The paper appropriately notes that “Because all responses were collected at one point, the models are estimating concurrent associations” and that a smaller administrative-support coefficient after including adjustment “can't establish that adjustment settles the effect of institutional support” . This is a strength.

However, elsewhere the prose risks drifting toward stronger substantive interpretation than the design can bear. For example, the Abstract states that academic adjustment “emerged as the stronger predictor of intended retention” and “explaining roughly 43% of the departure in students' responses” . The Results and Discussion repeatedly frame adjustment as central to retention in ways that may be read causally by non-specialist readers . The manuscript should standardize its language around association, not effect or influence, unless a stronger design is used.

6. The manuscript’s evidentiary contribution is modest and should be positioned more conservatively

The Introduction explicitly states that the study “does not introduce a new counselling intervention or a test of actual dropout” and instead “boundedly contributes” via perceived counselling administration, intended retention, and a measurement audit . I agree with that narrower framing. The problem is that later sections sometimes imply a broader administrative significance than the data justify, especially given reliance on self-reports, intentions rather than actual persistence, one institution, and n=65 for the main analyses .

The paper would be stronger if it explicitly presented itself as a preliminary institutional study or pilot measurement-focused investigation rather than a more general empirical account of student retention dynamics.

7. Reporting transparency is incomplete for reproducibility

The analytic protocol reports ordinary least squares models with HC3 robust standard errors and mentions demographic-adjusted sensitivity analyses using sex and age bands . Regression tables are provided with coefficients, confidence intervals, p-values, and R² . This is useful, but not sufficient for full reproducibility.

The manuscript should report:

• the full questionnaire items verbatim
• coding direction for each item
• missing-data handling
• whether assumptions beyond heteroskedasticity were checked
• exact model formulas
• whether age was treated categorically or ordinally in sensitivity analyses
• recruitment dates
• response completion criteria
• institutional ethics approval or exemption status

I did not find a clear ethics statement in the retrieved text. Given that human participants were surveyed about educational experiences via online forms, ethics/consent procedures should be explicitly documented in the manuscript.

8. Possible paper-mill or AI-assisted drafting traits warrant caution

I am obligated to flag that the manuscript displays multiple features often associated with unreliable drafting pipelines: pervasive formulaic filler phrases (“in actual application,” “broadly speaking,” “to a significant degree,” “within this framework”), syntactic corruption, odd lexical substitutions (“papering,” “spotted,” “autonomous coefficient”), and citation entries with embedded nonsensical phrases in the reference list . These patterns do not prove fabrication, but they do raise integrity concerns about how the text was produced and whether references, paraphrases, and interpretations were carefully verified by the authors.

I do not see direct evidence of data fabrication in the extracted statistical reporting alone. The numerical story is internally plausible and the manuscript openly discusses weaknesses. But the text quality is sufficiently anomalous that I recommend heightened editorial scrutiny, including verification of source references, raw data availability, and questionnaire provenance.

Potential Minor Revisions

1. Typographic, grammatical, and wording issues

The manuscript contains extensive local errors. A non-exhaustive list follows.

These examples come directly from the manuscript text .

2. Formatting and presentation issues

• The title line appears crowded and contains a typo: “ADJUSTUSTMENT.”
• Punctuation is inconsistent, including doubled semicolons in the research questions: “intended retention;; also” .
• Several tables appear to have formatting or labeling irregularities, including awkward inserted prose in table titles such as “Table 3. Building on this observation, hC3-robust regression estimates...” and “Table 4. In this broader context, scoring sensitivity...” .
• Capitalization is inconsistent for cited authors, for example “adams (2023)” and “baker and Siryk (1984)” in running text .
• The reference list includes intrusive filler text inside citations, which is not acceptable scholarly formatting. For example: “Fortunato... (2026), on the whole; Counselling Psychology Quarterly...” and “Gyasi... (2025), to a significant degree...” .

3. AI content analysis for this post-2021 manuscript

Estimated percentage of AI-generated or AI-heavily-assisted prose: 70–90%.

Basis for this estimate:

• repeated discourse fillers at unnatural frequency
• semantic drift and lexical substitutions inconsistent with normal human revision
• sentence-level grammaticality that remains globally structured but locally corrupted
• contamination of the reference list with prose fragments
• homogeneous stylistic artifacts recurring across sections

Highlighted likely AI-affected sections:

• Abstract, especially phrases such as “this study inspected this relationship” and “explaining roughly 43% of the departure in students' responses”
• Introduction, including “intention to continue papering?” and “spotted as students”
• Methods and measures, including “second event of the statement,” “any administrative provision index,” and similar constructions
• Discussion and implications, where filler-driven inflation of simple claims is especially visible
• References, where intrusive filler phrases embedded inside bibliographic entries are particularly concerning

Assessed epistemic impact: The likely AI-assisted drafting does not automatically invalidate the quantitative findings, but here it has a substantial epistemic impact because it obscures construct definitions, weakens interpretive precision, and raises concerns about whether citations, paraphrases, and analytical descriptions were carefully checked. In this manuscript, the drafting issue is not cosmetic; it materially affects scientific credibility.

Recommendations

1. Rewrite the manuscript completely for clarity and scholarly English before resubmission. This should include the title, abstract, body text, tables, figure/table captions if any, and references.

2. Add a full appendix with all questionnaire items verbatim, their response scales, coding direction, and which items were excluded or retained in each composite.

3. Reframe the study explicitly as a preliminary, single-institution, cross-sectional association study based on self-reported intentions rather than actual retention behavior.

4. Reconsider the measurement model for administrative support. If GCAS is formative or descriptive, analyze it accordingly rather than presenting it as a quasi-scale with weak internal consistency.

5. Strengthen the Methods section by clarifying recruitment procedure, participant eligibility, duplicate-response safeguards, missing-data handling, sensitivity model specification, and ethics approval/consent procedures.

6. Tighten the inferential language throughout. Replace causal or quasi-causal wording with association-focused phrasing.

7. Verify every reference against the original sources and fully clean the reference list of intrusive non-bibliographic phrases.

8. If possible, provide supplementary material with anonymized data and analysis code or, at minimum, a detailed analysis workflow to improve reproducibility.

9. Consider collecting a larger, student-only sample across multiple institutions and using validated measures of adjustment and retention-related constructs in a future version of the study.

10. Before further submission, conduct an internal integrity audit of the manuscript text, questionnaire source, and reference management process. The current document contains enough linguistic anomalies to justify extra scrutiny.