Predatory publishing: detection, taxonomy, and the moving landscape

Last updated: 2026-05-20

Synthesis

“Predatory publishing” names the business model in which a venue charges article-processing fees while providing little or no editorial work, peer review, or indexing in return. The term became broadly current with Jeffrey Beall’s blog Scholarly Open Access, which from 2008 until its abrupt removal in 2017 maintained the most-cited list of “potential, possible, or probable” predatory journals and publishers [1]. Beall’s list was controversial in detail — accused of methodology issues, OA bias, and individual misclassifications — but the underlying phenomenon it documented is well-attested across the OA literature [2, 3]. Roughly two thirds of self-styled OA journals in the early 2010s were estimated to be either definitively predatory or operating without the quality controls a legitimate journal would maintain [2]; that fraction has shifted as the legitimate OA infrastructure matured, but the predatory landscape has matured too, and detection remains an active problem.

Several patterns dominate the predatory operating model:

• Pay-to-publish with no real peer review. Submissions are accepted with no substantive revision or external review, often within days. The fee is the product.
• Fabricated editorial boards. Editorial lists frequently include real scholars who never agreed to serve, sometimes with affiliations the listed scholar never held.
• Fake or misleading metrics. Journals invent impact factor analogues (“Universal Impact Factor”, “Global Impact Factor”), cite themselves to inflate Google Scholar h-index, or claim indexing in databases they’re not in [4].
• Hijacked journals. Established journal titles are cloned at look-alike URLs to harvest fees from authors searching for the legitimate journal [5].
• Aggressive solicitation. Mass-emailed calls for papers targeting researchers who have published nearby, often with flattering personalisation lifted from public CVs.

For an automated tool like scriptorium, the consequence is twofold. First, the predatory layer cannot be ignored — any venue recommendation that doesn’t filter for it is incomplete and exposes the author to a real cost. Second, detection is hard in-band and depends on community-maintained authoritative sources. A tool that pretends to verify predatory status from training-data knowledge alone will fail; the correct posture is to apply published heuristics, defer to authoritative checklists (Think.Check.Submit, COPE, DOAJ membership, Cabell’s Predatory Reports where the author has access), and refuse to recommend anything the heuristics flag.

Evidence and frameworks

Beall’s list: history and what it left behind

Beall’s blog [1] ran from 2008 to January 2017, at which point it was taken down without public explanation. Speculation centred on legal pressure from listed publishers; Beall himself later wrote about the experience in Biochemia Medica [6]. The blog’s specific recommendations are no longer available at the original URL but are archived; the methodology Beall used has been critiqued in detail [7, 8].

The most durable contribution of Beall’s work was not the list itself but the publication of detection heuristics. The criteria Beall published (lack of editorial board, missing peer- review process, fake metrics, unprofessional websites, predatory solicitation language) became the reference taxonomy for subsequent work. The criteria themselves are not in serious dispute even when individual judgments using them were. A modern detection effort still uses Beall’s framework, just with more maintained data sources.

Successor resources

Several community and commercial resources have filled the gap since 2017:

Cabell’s Predatory Reports [9] is the commercial successor most research libraries subscribe to. It maintains both a “Journal Whitelist” (positively reviewed venues) and “Journal Blacklist” (predatory or otherwise problematic venues), with documented criteria for each listing. Access requires an institutional subscription, which limits its reach for unaffiliated researchers but makes it the most authoritative single source where access is available.

Think.Check.Submit [10] is the community-maintained checklist many editorial guides now point at. It does not maintain a list of journals; it provides a structured checklist of questions an author can answer about any specific venue before submitting. The checklist covers editorial board verification, peer-review process transparency, indexing claims, society affiliation, and contact information consistency. Think.Check.Submit is the right resource to recommend to authors as the process they should follow, even when an automated tool has not flagged a specific concern.

COPE (Committee on Publication Ethics) [11] maintains a membership-based set of publishing-ethics standards. COPE membership is a positive signal but not a certification — predatory publishers have applied for and even achieved COPE listings before being removed, and many legitimate journals are not COPE members because of cost or institutional structure.

DOAJ (Directory of Open Access Journals) [12] maintains a curated list of OA journals meeting transparency and quality criteria. DOAJ inclusion is a moderately strong positive signal; DOAJ removal (the journal was once listed and then excluded) is a strong negative signal worth surfacing if visible. The DOAJ Seal indicator marks journals meeting additional rigor criteria.

OASPA (Open Access Scholarly Publishers Association) [13] membership requires meeting publishing-practice standards including peer-review documentation and editorial-board transparency. OASPA membership is a positive signal for OA publishers similar to DOAJ inclusion for journals.

For an automated venue-fit recommendation, the right posture is to use these resources as soft signals — DOAJ inclusion is a positive heuristic, Cabell’s flagging is a negative heuristic — but not to treat any single source as definitive. Predatory designation is contested in detail and an automated tool should defer to authoritative human-curated sources rather than rendering its own judgments.

Per-journal, not per-publisher: the MDPI nuance

A common error in informal predatory-detection lists is treating entire publishers as predatory based on the worst journals in their portfolio. The most cited contemporary case is MDPI, a Basel-headquartered OA publisher with several hundred journals spanning quality levels. MDPI as a publisher has been controversial for years — high APCs, aggressive solicitation, fast turnaround, and self-citation patterns drew scrutiny [14, 15] — but the publisher includes journals that have built solid reputations in their fields (Sensors, Cells, International Journal of Molecular Sciences, Cancers) alongside journals that have raised serious quality concerns. Treating MDPI as a single entity is wrong in both directions: it would flag legitimate venues and it would miss problems in specific journals within the portfolio.

The principle: predatory assessment must be per-journal, not per-publisher. The same publisher can host both a venue with rigorous review and a venue with predatory characteristics. The relevant question is always about the specific journal under consideration, with the publisher as one signal among several.

The Hindawi-Wiley case: when legitimate publishers degrade

In 2023-2024, Wiley closed 19 Hindawi-acquired journals after documented manipulation by paper mills [16, 17]. Hindawi had been a long-standing OA publisher acquired by Wiley in 2021; subsequent investigation found that several of its journals had been heavily compromised by special-issue submission patterns correlating with paper-mill operations. The case is significant for venue-fit work because it demonstrates that legitimacy is not permanent: a journal can move from defensible to compromised within a few years, especially in the OA-special-issue model where revenue scales with publication volume and editorial oversight can lag.

Practical consequence: a venue-fit recommendation based on training-data knowledge of a journal’s status from before the events in question is potentially stale. The skill must caveat that its knowledge has a date stamp and that authors should verify current status via Think.Check.Submit or DOAJ at submission time.

Hijacked journals

A distinct subcategory worth naming: hijacked journals, where a predatory operation creates a look-alike domain or even a full clone of a legitimate journal’s website to intercept submissions and fees [5]. The legitimate journal still exists at its real URL; authors searching the journal name and clicking the wrong result end up paying APCs to the hijacked version, which provides no actual publication or indexing.

Detection requires URL verification rather than name verification. A skill recommending venues should provide the verified URL for each recommendation and flag any case where multiple URLs are associated with a single journal name (a common signal of hijacking in progress).

How this informs scriptorium

For venue-fit specifically:

1. Predatory refusal is a load-bearing posture, not an aside. No recommendation list includes a flagged venue. The skill maintains a ## Predatory signals detected section in its output even when no candidates were flagged, stating explicitly that the check was applied — silence here is indistinguishable from “we didn’t check”, which is the wrong inference for the author to draw.

2. Use per-journal, not per-publisher, judgments. The skill does not flag a publisher as predatory; it flags a specific journal as predatory if the heuristics fire. MDPI and similar mixed-portfolio publishers require per-journal assessment.

3. Defer to authoritative human-curated resources where they exist. Think.Check.Submit, DOAJ, OASPA, and Cabell’s are the references. The skill applies published Beall-style heuristics and, when uncertain, says so and points the author at the authoritative source rather than guessing.

4. Caveat staleness. Training-data knowledge of a venue’s status is potentially out of date. The skill’s recommendation output names this explicitly: “venue status assessed from training data through [model knowledge cutoff date]; verify current status via Think.Check.Submit at submission.”

5. Hijacked-journal protection via URL verification. When recommending a venue, the skill provides a verified URL and flags multiple-URL situations. The recommendation includes the verified URL even when the author hasn’t asked, because the missing-URL failure mode is silent.

6. Refuse cleanly at the boundary. When the author asks whether a specific journal is predatory, the skill answers based on the heuristics it can apply. When the heuristics are inconclusive, it says so, names the heuristics it checked, and points the author at Think.Check.Submit for the remaining verification steps. It does not invent a verdict.

Implementation priority for scriptorium

Verdict: Direct grounding for the venue-fit skill (v0.2). The predatory-publishing dimension is first-class, not an aside. Without this grounding the skill would be incomplete — the most expensive misfit failure mode for an author is publishing in a predatory venue, and not flagging that is worse than not recommending at all.

Why useful context anyway:

• The per-journal-not-per-publisher principle applies to any future scriptorium work that assesses venues (e.g., a hypothetical preprint-server fit skill would face the same question about preprint-server moderation rigor varying within a publisher).

• The staleness caveat — “training-data knowledge has a date stamp; verify current status” — is the same shape as the citation-audit caveat about CSL metadata not standing in for full-text verification. It’s a principle: an LLM-driven tool must be honest about what its training data can and can’t attest to about the current world.

• The “use authoritative human-curated sources” pattern is duplicated across multiple knowledge notes in this layer (predatory-publishing for venues, [[citation-claim- alignment]] for full-text verification, [[reporting- guidelines]] for checklists). Worth being explicit that this is a project-wide pattern: scriptorium is the assistant on the edge of the workflow; authoritative sources of truth remain external.

Condition that would flip this: if a maintained, machine- queryable predatory-detection API emerges (Cabell’s API access at scale, a successor to Beall’s list with API endpoints), the skill could shift from heuristic-based flagging to API-driven flagging. That’s a future capability, not a v0.2 dependency.

Cross-references

• venue-selection — venue-fit’s primary grounding; predatory refusal is the layer applied across every recommendation that note produces.
• editorial-decision-making — desk-rejection rates and editor triage; predatory venues are by construction not in this framework.
• reference-managers — adjacent territory; reference managers face the parallel “is this citation real” problem that predatory journals create downstream of acceptance there.
• declared-work-scope — the convention. Predatory venue refusal is consistent with the principle: scriptorium operates on declared work and refuses to recommend venues that would exploit that work.

References

[1] Beall, J. (2008-2017). Scholarly Open Access. Blog. Original URL no longer active (removed 2017-01-15); archived at the Internet Archive and Wayback Machine.

[2] Shen, C., & Björk, B.-C. (2015). ‘Predatory’ open access: a longitudinal study of article volumes and market characteristics. BMC Medicine, 13, 230. DOI: 10.1186/s12916-015-0469-2. PMID: 26423063. (The empirical estimate of predatory OA scale in the early-to-mid 2010s.)

[3] Kakamad, F. H., Mohammed, S. H., Najar, K. A., Qadr, G. A., Ahmed, J. O., Mohammed, K. K., Salih, R. Q., Hassan, M. N., Mikael, T. M., Kakamad, S. H., Baba, H. O., Salih, A. M., Othman, S., & Ahmed, M. S. (2019). Kscien’s list; a new strategy to discourage predatory journals and publishers. International Journal of Surgery Open, 17, 5-7. DOI: 10.1016/j.ijso.2019.01.002. [TODO verify exact volume/page; the Kscien list itself is the more durable reference here than the methodology paper.]

[4] Jalalian, M., & Mahboobi, H. (2014). Hijacked journals and predatory publishers: Is there a need to re-think how to assess the quality of academic research? Walailak Journal of Science and Technology, 11(5), 389-394. (Early documentation of the “hijacked journals” pattern.)

[5] Dadkhah, M., Maliszewski, T., & Teixeira da Silva, J. A. (2016). Hijacked journals, hijacked web-sites, journal phishing, misleading metrics, and predatory publishing: actual and potential threats to academic integrity and publishing ethics. Forensic Science, Medicine, and Pathology, 12(3), 353-362. DOI: 10.1007/s12024-016-9785-x. PMID: 27342770. (More extensive taxonomy of hijacking and metric-faking patterns.)

[6] Beall, J. (2017). What I learned from predatory publishers. Biochemia Medica, 27(2), 273-278. DOI: 10.11613/BM.2017.029. PMID: 28694718. (Beall’s own post-shut- down reflection on the methodology and the pressures.)

[7] Olivarez, J. D., Bales, S., Sare, L., & vanDuinkerken, W. (2018). Format aside: Applying Beall’s criteria to assess the predatory nature of both OA and non-OA library and information science journals. College & Research Libraries, 79(1), 52-67. DOI: 10.5860/crl.79.1.52. (Independent application of Beall’s criteria to non-OA journals — finds the criteria flag problematic non-OA venues too, suggesting the criteria are about quality, not OA-ness per se.)

[8] Berger, M., & Cirasella, J. (2015). Beyond Beall’s list: Better understanding predatory publishers. College & Research Libraries News, 76(3), 132-135. DOI: 10.5860/crln.76.3.9277. (Critique of Beall’s methodology and the OA-vs-predatory conflation; widely cited.)

[9] Cabell’s International. (Ongoing.) Predatory Reports (formerly Cabell’s Blacklist). https://www2.cabells.com. Subscription required. (Commercial successor to Beall’s list; the most authoritative single source where access is available.)

[10] Think.Check.Submit. (Ongoing.) Choose the right journal for your research. https://thinkchecksubmit.org. (Community- maintained checklist; the recommended process resource for authors. Funded jointly by COPE, DOAJ, OASPA, and others.)

[11] Committee on Publication Ethics (COPE). (Ongoing.) COPE Core Practices. https://publicationethics.org. (Membership- based publishing ethics standards.)

[12] Directory of Open Access Journals (DOAJ). (Ongoing.) https://doaj.org. (Curated OA journal directory; DOAJ Seal indicates additional rigor criteria.)

[13] Open Access Scholarly Publishers Association (OASPA). (Ongoing.) https://oaspa.org. (Membership organisation for OA publishers; membership criteria publicly listed.)

[14] Crosetto, P. (2021). Is MDPI a predatory publisher? Blog post. <https://paolocrosetto.wordpress.com/2021/04/12/ is-mdpi-a-predatory-publisher/>. (Influential informal analysis of MDPI’s growth and self-citation patterns. Conclusion: publisher-level designation is wrong, but specific concerns about specific journals within the portfolio are real. The reference is a blog post rather than a peer-reviewed paper; cited here because it is the most widely-referenced contemporary discussion of the MDPI question.)

[15] Petrou, C. (2022). Guest Post — MDPI’s Remarkable Growth. The Scholarly Kitchen. <https://scholarlykitchen.sspnet.org/ 2020/08/10/guest-post-mdpis-remarkable-growth/>. [TODO verify date; the URL slug suggests 2020 not 2022.] (Industry-side analysis of MDPI’s growth and the structural questions it raised about the OA-special-issue model.)

[16] Else, H. (2024). Paper-mill detector put to the test in biggest-ever publishing-fraud sweep. Nature, 627, 261-262. DOI: 10.1038/d41586-024-00580-0. (Coverage of the Wiley-Hindawi shutdown and the paper-mill detection work behind it.)

[17] Brainard, J. (2023). Fake scientific papers are alarmingly common. Science, 380(6645), 568-569. DOI: 10.1126/science.adi9617. (Broader context on paper-mill operations and their effect on OA-venue integrity.)