Paper mills are increasingly targeting engineering journals with manuscripts that appear technically sound but lack real scientific integrity. For editors, the challenge is no longer identifying poor-quality papers, but detecting sophisticated submissions designed to pass initial screening.
This article provides a practical, engineering-focused framework to identify paper mill submissions quickly, using clear editorial signals, methodological checks, and reproducibility criteria.
1. What “paper mills” look like in engineering
Paper mills produce manuscripts at scale using templates, recycled data, and fabricated authorship. In engineering, they tend to target areas where validation is easier to fake, such as applied AI, IoT systems, materials with “novel” compositions, or control systems validated only through simulation.
These papers are designed to look credible at first glance. The problem is not formatting or language—it is the lack of real scientific coherence behind the text.
2. Fast editorial triage (≤10 minutes)
Before sending a manuscript to review, a quick screening can filter out a large proportion of suspicious submissions.
Look at scope and positioning. Titles are often generic (“A Novel Method for…”) and fail to define a concrete engineering problem. Keywords may feel artificially aligned with trending topics rather than the journal’s scope.
Check the structure. These papers are usually perfectly formatted but shallow. Sections exist, yet the connection between methods and results is weak.
Assess content coherence. The problem statement is often vague or exaggerated, and conclusions tend to overclaim what the results actually support.
If several of these issues appear together, the manuscript is likely not suitable for peer review.
3. Textual and linguistic signals
Paper mill manuscripts are frequently generated or semi-generated, which creates recognizable patterns.
You may notice repeated phrases across sections, inconsistent terminology (the same variable renamed multiple times), or sentences that are grammatically correct but scientifically empty. Another common issue is abrupt changes in writing style, suggesting multiple sources stitched together.
A typical example is a statement like:
“The experimental results verify the effectiveness of the proposed method”
without any quantitative evidence.
4. Figures, tables, and data anomalies
Visual elements are often one of the weakest points.
Figures tend to be overly clean: smooth curves without noise, unclear units, or generic plotting styles reused across unrelated topics. Resolution and labeling may also be inconsistent.
Tables often show unrealistic precision, such as identical decimal patterns across metrics, and frequently lack any error analysis or variance reporting. Benchmark comparisons may be included without properly identifying datasets or references.
A key warning sign is inconsistency between sections—for example, methods describing one experiment while results present something different. Lack of traceability (no dataset, no code, no parameter details) is especially critical.
5. Methodological red flags (engineering-specific)
In engineering, weak methodology is often easier to detect than textual issues.
Common problems include absence of meaningful baselines, reliance on simulation without real-world constraints, and missing details about parameter tuning. There is also frequent overuse of buzzwords—AI, blockchain, digital twin—without any real implementation depth.
A simple test is useful here: could a competent reviewer reproduce the work using the information provided? If the answer is no, the manuscript does not meet minimum scientific standards.
6. Citation patterns
References can reveal systematic issues.
Suspicious manuscripts often include clusters of unrelated citations, excessive self-citation, or references to obscure or irrelevant journals. In some cases, cited works do not actually support the claims made in the text.
It is worth checking whether key foundational papers in the field are missing, or whether cited methods are not actually used in the study.
7. Authorship and metadata signals
Metadata inconsistencies are another strong indicator.
Email addresses may not match institutional domains, affiliations can be difficult to verify, and author profiles may appear across multiple unrelated topics. Some patterns include repeated submissions from the same group with minimal variation in titles or content.
These signals alone are not definitive, but combined with technical issues they become highly relevant.
8. Cross-paper similarity detection
Even without specialized software, patterns can be identified across submissions.
Look for identical structures, reused figures with minor modifications, or introductions that are nearly the same but with keywords swapped. Searching specific phrases in Google Scholar can sometimes reveal parallel or duplicated submissions.
9. What to do when you detect a paper mill
If suspicion is strong, the manuscript should not proceed to peer review.
A desk rejection is appropriate. It is better to keep the decision concise and avoid detailed feedback, as extensive comments can help paper mills refine future submissions. Internally, it is useful to track recurring names, emails, or patterns. If the issue repeats, informing the publisher may be necessary.
In borderline cases, requesting raw data or code can help clarify the situation.
10. Minimal rejection template
A simple and neutral response is sufficient:
“The manuscript does not meet the journal’s standards in terms of methodological transparency and scientific robustness. Therefore, it will not be considered for peer review.”
11. Key principle
Paper mills optimize for appearance, not for scientific integrity.
The goal at the editorial stage is not to prove misconduct, but to identify lack of reproducibility, insufficient technical depth, and internal inconsistency. When those elements are missing, rejection is already justified.