Preprint landscape: servers, post-publication review, and the moving boundary

Last updated: 2026-10-20

Synthesis

The preprint ecosystem has expanded from a narrow physics-and-CS convention (arXiv, founded 1991) into a discipline-spanning infrastructure with substantial — and still-evolving — overlap with traditional journal publishing. By 2024, biomedical preprint servers were hosting hundreds of thousands of papers annually; COVID-19 (2020-2022) compressed roughly a decade of biomedical preprint normalisation into two years [1, 2]. The current landscape encompasses archive-style preprint servers that provide minimal moderation and lasting time-stamped citation (arXiv, bioRxiv, medRxiv, ChemRxiv, EarthArXiv, PsyArXiv, SSRN, OSF Preprints), post-publication peer-review platforms that add a refereed layer on top of preprints (PCI / Peer Community In, Review Commons, F1000Research, eLife’s post-2022 “reviewed preprints” model), and journal-policy variation about preprints that ranges from required pre-submission to outright prohibition [3, 4].

Since 2025 the infrastructure itself has been reorganised, and funders have started to expect or require preprints. bioRxiv and medRxiv moved to an independent nonprofit, openRxiv, in March 2025 [22]. arXiv left Cornell to become an independent nonprofit in 2026 [5]. The Gates Foundation has expected funded manuscripts to be posted as preprints since January 2025, “to the extent feasible” and with an exception where the grantee judges a preprint inappropriate [27]. HHMI has required initial and revised preprints on designated servers since January 2026, for original research articles that are major contributions from covered HHMI labs [26]. Under submission pressure, partly from AI-generated papers, arXiv also tightened what it accepts. Since October 2025 its CS category requires review and position papers to have passed peer review first [21], and since October 2026 every submitter is limited to two submissions per calendar month [20].

The decision an author faces is no longer just “where do I submit?” but a multi-part one: do I preprint? where? before, after, or instead of journal submission? do I want pre- or post- publication peer review? what does my target journal allow? These questions have substantively different answers across fields (physics and CS have preprint cultures decades older than the journal venues that aggregate them; clinical medicine still treats preprinting cautiously because of patient-decision implications), across career stages (preprinting carries different cost-benefit calculations for trainees vs. established researchers), and across funder requirements (NIH, Wellcome Trust, cOAlition S funders all have specific preprint and open-access requirements that interact with venue choice).

For an automated venue-fit recommendation, the consequence is that the preprint layer cannot be a one-size answer. The author needs to be asked whether preprints are in scope for them, and the recommendation has to account for the chosen preprint server’s interaction with the recommended journal’s preprint policy. The landscape moves fast enough that the right grounding posture is to maintain principles in this note and verify specifics at invocation time.

Evidence and frameworks

The major preprint servers

arXiv was founded in 1991 by Paul Ginsparg, hosted at Cornell for decades, and established as an independent nonprofit in 2026 [5]. It is the original and the model. Initially physics, expanded to mathematics, computer science, quantitative biology, economics, statistics, and several others. arXiv applies endorsement-based moderation (new authors must be endorsed by an existing author) and minimal subject-area screening; the time-stamp is the durable contribution. arXiv now reports hosting more than three million articles in eight subject areas [5]. Volume is rising fast: arXiv received 40,363 submissions in September 2026, and submissions doubled in the two years before that [20]. The CS subfields have built an entire publication culture around arXiv-first posting; physics and math have done so for longer.

Two 2025–2026 arXiv policy changes matter for venue advice:

• CS review and position papers. Since late October 2025 (arXiv blog post of 31 October 2025), review (survey) articles and position papers in arXiv’s CS category must already be accepted at a journal or conference, after completed peer review, before arXiv will consider them. Workshop review is “not enough”. arXiv cited an “unmanageable influx” of such papers, many of them “little more than annotated bibliographies”, with generative AI adding to the flood. arXiv frames this as enforcing its existing rules on accepted content types rather than as a new policy [21].
• Rate limits. From 1 October 2026 each submitter may make at most two submissions per calendar month, across all categories. The separate cap of three active submissions at any time has applied since 2024. Rejected submissions count toward the monthly limit. arXiv calls this a “stopgap” while it works out best practice for authors using AI tools [20].

bioRxiv and medRxiv began as projects of Cold Spring Harbor Laboratory; since March 2025 they have been overseen by openRxiv, an independent nonprofit with researcher-led governance and initial funding from the Chan Zuckerberg Initiative [22]. bioRxiv (launched 2013) covers preclinical and basic biology; medRxiv (launched 2019, founded by CSHL, Yale and BMJ [4, 6]) covers clinical research. Both apply a screening process focused on excluding non-research content, plagiarism, and obvious methodological problems. medRxiv adds a clinical-screen layer to guard against papers that could mislead patient decisions [6]. COVID-19 made medRxiv a central piece of the public-health information infrastructure and surfaced both the value (rapid dissemination of preliminary findings) and the cost (misinterpreted preprints driving public policy) of biomedical preprinting [7].

ChemRxiv (founded 2017, jointly run by ACS, RSC, GDCh, CCS, and CSJ) is the chemistry-discipline preprint server, with moderation similar to bioRxiv’s. EarthArXiv (2017, EarthRxiv governance) covers Earth and planetary sciences. PsyArXiv (2016, hosted on OSF) is the psychology preprint server with broader social-sciences uptake. SSRN (1994, acquired by Elsevier in May 2016 [8]) is the social sciences and economics preprint server with substantial uptake in finance, law, and accounting.

OSF Preprints (Open Science Framework, Center for Open Science) provides a multi-disciplinary preprint platform with moderation per discipline. The OSF model is more meta-infrastructure than single server; many discipline-specific preprint servers run on OSF (PsyArXiv, SocArXiv, LawArXiv, others).

For an author choosing a preprint server, the relevant axes are: discipline scope (does this server cover your field?), moderation rigor (will format/scope problems be caught before your paper is publicly time-stamped?), license and reuse terms (CC-BY, CC-BY-NC, and variants — these matter downstream for journal submission), indexing (Google Scholar yes for all major servers; PubMed-relatedness varies), and the target journal’s preprint policy. A further axis is now funder designation. For articles its policy covers, HHMI requires bioRxiv (or medRxiv, ChemRxiv or arXiv for out-of-scope work) under CC BY 4.0 and says “other preprint servers should not be used” [26]. Where Gates-funded authors preprint, the policy asks for a server “recognized by the foundation or preapproved preprint server which applies a sufficient level of scrutiny to submissions” [27]. For funded authors, server choice can be a compliance question rather than a preference.

Pre-publication vs post-publication peer review

The pre/post-pub distinction is the second axis. Traditional journal review is pre-publication: a manuscript is reviewed before it appears in the venue. Post-publication review platforms add an explicit layer on top of preprints in which the time-stamped preprint is the durable record and review happens after public posting. Several variants matter:

PCI / Peer Community In [9] is a community-run system in which discipline-specific “Peer Communities” (PCI Evolutionary Biology, PCI Ecology, PCI Neuroscience, others) recommend preprints based on full peer review. The recommended preprint is a recognised endorsement; no APC, no journal submission required. Many mainstream journals accept PCI-recommended preprints as fast-tracked submissions or accept them outright [10]. PCI is strongest in ecology and evolutionary biology and is expanding.

Review Commons [11] is a refereed-preprint service that provides journal-independent peer review on bioRxiv preprints, with the resulting reviews portable to affiliated journals (EMBO Press, eLife, others). The author receives reviews; the journals receive reviews and decide independently whether to accept based on the existing review record. The model is “review the science once, decide at multiple journals” — the inefficiency of multiple re-review cycles is reduced.

F1000Research [12] is a post-publication peer-review platform where papers are published openly after editorial check (not peer review), then reviewed openly by invited reviewers whose names and reports are public. Articles can be revised in response to reviews; the final status of an article is “approved” (1+ reviewer recommendations), “approved with reservations”, or “not approved”, visible to readers. F1000Research has strong uptake in clinical research, gene/genome papers, software/method papers, and is the canonical reference for the post-pub review model.

eLife’s post-2022 model [13] is the most consequential recent shift. eLife announced in October 2022 that it would eliminate accept/reject decisions after peer review [13]; under the current model, eLife reviews any preprint the editors select for review and publishes the resulting Public Reviews alongside the preprint as a “reviewed preprint”, with no final acceptance gate. The author can revise; eLife continues to review revisions; the preprint and its review record are the publication. This is the most aggressive existing implementation of the post-publication-review model at a high-impact journal. It has an indexing cost. Web of Science put eLife on hold in October 2024. eLife now sends Web of Science only the papers assessed as “solid” or better, about 93% of them, and is therefore not eligible for a Journal Impact Factor. In Scopus, eLife moved to the Preprints Collection. eLife reported in May 2025 that more than 100 funders and institutions still consider eLife papers in research evaluation [23].

The pre/post-pub choice is not just a venue choice; it is a strategic choice about timing (rapid dissemination vs. embargo), transparency (named open reviews vs. blinded journal reviews), community signal (a PCI recommendation carries different weight than a journal acceptance, and the weight varies by field), and CV interpretation (search and tenure committees treat reviewed preprints inconsistently; varies by field and institution).

Journal preprint policies

Journal policies on preprints range from required to prohibited, with most journals at “permitted with conditions” in the middle. The variation matters because some authors have been desk-rejected for either preprinting (some journals consider this prior publication) or not preprinting (some funders or journals now require open-access pre-submission posting). The authoritative reference for journal-by-journal policies is SHERPA/RoMEO [14] (now JISC’s Open Policy Finder), which aggregates publisher self-disclosed policies on preprint, accepted manuscript, and published version self-archiving.

Common policy patterns:

• Preprint-friendly (most life sciences journals as of 2024): preprinting before submission is permitted with the requirement that the manuscript be appropriately cited in the published version and on the preprint server once accepted. Examples: eLife, PLOS Biology, Nature journals, Cell Press journals (with stipulations).

• Preprint-required (an emerging category): preprinting at a specified server is a precondition for submission or consideration. eLife’s post-2022 model is the clearest example; EMBO Press journals (some) accept bioRxiv-posted preprints directly for review. Funder policies now push the same way from the author side: HHMI requires preprints for covered original research articles from 2026 [26], and Gates has expected them, to the extent feasible and with grantee exceptions, since 2025 [27].

• Preprint-tolerant (most clinical and applied-medicine journals as of 2024): preprinting is permitted but the journal treats it as a minor demerit in some cases, especially when the preprint received public attention before review.

• Preprint-prohibited or hostile (a shrinking but persistent category): some specialty journals, particularly in clinical surgery and some humanities subfields, treat preprinting as prior publication and decline to consider preprinted work. The list of such journals shifts; SHERPA/RoMEO is the reference.

For a venue-fit skill, the practical pattern is to check the recommended journal’s preprint policy against the author’s preprint preference and surface any incompatibility.

Funder requirements

Funder requirements have been a strong recent driver of preprint adoption. The cOAlition S “Plan S” [15] requires deposit in an open repository for funded research. NIH’s 2024 Public Access Policy took effect on 1 July 2025, brought forward from 31 December 2025. It requires the Author Accepted Manuscript to be submitted to PubMed Central upon acceptance and made publicly available without embargo on the official date of publication [16]. Separately, in July 2025 NIH proposed five options for limiting allowable publication costs. They ranged from disallowing all publication costs to per-article caps of $2,000–$6,000 and an award cap of 0.8% of direct costs or $20,000, whichever is greater, with a proposed effective date of 1 January 2026 [24]. The most recent dated status review found (March 2026) reported that no cap had been finalised. A check of NIH’s list of policy notices in October 2026 found no notice finalising one either [24]. Wellcome Trust funds APCs only for research articles in fully open-access journals or platforms indexed in DOAJ, from 1 January 2025. It strongly encourages preprints, requires them during public health emergencies, and accepts a CC BY preprint as compliant where neither the Version of Record nor the Author Accepted Manuscript can be made open access [25]. Howard Hughes Medical Institute (HHMI) has the most prescriptive policy. For articles submitted from 1 January 2026, it requires initial and revised preprints under CC BY 4.0 on a designated server (bioRxiv preferred). The requirement covers named HHMI roles (Investigators, Janelia group leaders, Freeman Hrabowski Scholars and others listed in the policy), and only original research articles that are major contributions from their labs: someone from the lab is (co-)first, (co-)last or (co-)corresponding author. It “does not apply to review articles, reviews of existing protocols and methods, perspectives, commentaries” [26]. The initial preprint must go up promptly after a journal sends the manuscript to review; HHMI strongly recommends posting by journal submission. The revised preprint must go up before journal publication. HHMI budgets can no longer pay APCs at hybrid journals [17, 26]. The Gates Foundation’s refreshed policy has applied to all funded manuscripts since 1 January 2025. It states the foundation’s “expectation” that funded manuscripts are shared promptly as preprints, “as soon as possible and to the extent feasible”, “except where the grantee determines that a preprint is not appropriate due to ethical, safety or other legitimate concerns”. It requires immediate CC BY 4.0 availability, and states that the foundation “will not pay article processing charges” [27]. Preprinting is therefore a strong default for Gates grantees, not a hard requirement.

For an automated venue-fit recommendation, funder-driven constraints are hard filters: a Wellcome-funded researcher recommended to submit to a subscription journal without an OA option is recommended to violate their funder agreement. Hybrid journals are now a funder issue too. Wellcome and HHMI no longer pay hybrid APCs [25, 26], and Gates pays no APCs at all [27]. The skill should ask about funder constraints up front when the manuscript’s financial scaffolding is non-trivial.

What the landscape says about itself: known unknowns

The preprint landscape moves faster than peer-reviewed evidence about the preprint landscape moves. Several questions are actively unresolved as of late 2024 / early 2026:

• Does preprinting affect citation rates and journal acceptance rates? Multiple studies show modest positive effects on citation [18, 19], but the cleanest causal evidence is still thin and the publication-bias confound is real.

• How do search and tenure committees weight preprints? Cross-institutional variation is large; this is a CV-strategy question without a clean answer.

• Will pre-publication review survive in fields that adopt reviewed-preprint models? eLife’s bet is that the answer is no; the broader journal ecosystem’s response is in progress.

• What is the right policy on AI-generated preprints? Several servers have begun requiring disclosure; standards are not uniform. arXiv’s responses so far are restrictions on content type and volume, not disclosure rules: the CS review/position-paper rule [21] and the 2026 rate limit [20].

For an automated tool, the operational consequence is that this note should be treated as guidance to principles for recommendation, with the specifics to be verified at invocation time. The author should be pointed at SHERPA/RoMEO, DOAJ, and the specific journal’s instructions when the decision is being made, not handed an authoritative answer from training data alone.

How this informs scriptorium

For venue-fit specifically:

1. Preprints are an opt-in mode, asked once. The skill asks whether preprint recommendations are in scope before producing them. Default depends on field (life sciences and CS / physics: default ask; clinical surgery, some humanities: default ask with caveat that this is field-uncommon). When in scope, the preprint section is its own tier (not interleaved with journals); when out of scope, it’s suppressed entirely.

2. Recommend specific preprint servers, not “preprint it somewhere”. Match the server to the manuscript’s discipline (bioRxiv for life sciences, medRxiv for clinical, arXiv for physics/CS, ChemRxiv for chemistry, OSF for cross-discipline). Surface the per-server moderation rigor, license, and indexing so the author sees the trade-offs. Do not recommend arXiv CS for an unreviewed survey or position paper [21]. For HHMI-covered original research articles, restrict the choice to HHMI’s designated servers [26].

3. Surface the pre/post-pub review choice as a strategic decision. Within the preprint section, name PCI, Review Commons, F1000Research, and eLife’s reviewed-preprint model as the strategic alternatives to traditional review. The choice involves timing, transparency, community signal, and field convention — not a single “best” answer.

4. Check journal preprint policy against preprint recommendation. When recommending both a journal and a preprint server, verify (or caveat) that the journal accepts preprinted work from that server. Flag any incompatibility.

5. Check funder constraints. When the manuscript declares funder requirements (NIH, Wellcome, cOAlition S, etc.), filter the recommendation list against those requirements before tiering. A Wellcome paper recommended to a subscription journal without an OA option is a wrong recommendation. HHMI is the one funder here whose preprint rule is a requirement: for an original research article that is a major contribution from a covered HHMI lab ((co-)first, (co-)last or (co-)corresponding author), preprinting moves from opt-in to required; reviews, perspectives and commentaries are exempt [26]. A Gates-funded manuscript should get a strong prompt to preprint, not an enforced requirement, because the policy is an expectation “to the extent feasible” with grantee-determined exceptions [27].

6. Caveat staleness, defer to SHERPA/RoMEO at decision time. The skill names that its preprint-server and journal-policy knowledge has a date stamp and points the author at SHERPA/RoMEO for verification before submission. The verification step is part of the recommended action sequence, not an aside.

Implementation priority for scriptorium

Verdict: Direct grounding for the venue-fit skill (v0.2); preprints are an explicit mode within that skill, not a separate skill. The same-skill decision is documented in issue #82.

2026-10-20 refresh: the verdict holds, but two behaviours need updating in the skill (tracked for the skills pass). First, a declared HHMI funder should turn the preprint mode from opt-in into required, but only for original research articles that are major contributions from a covered HHMI lab, and should restrict server choice [26]. A declared Gates funder should trigger a strong prompt to preprint, not a requirement, because Gates expects preprints “to the extent feasible” and lets grantees decide a preprint is inappropriate [27]. Second, the server-matching rule should exclude arXiv CS for unreviewed review and position papers [21]. Funder APC rules (Wellcome DOAJ-only, HHMI no hybrid, Gates none, NIH caps proposed but not finalised as of October 2026) belong in the hard-filter step [24–27].

Why useful context anyway:

• The opt-in preprint mode pattern (default ask, then proceed) is the right shape for any decision-support feature that bridges fields with different cultural defaults. Worth being explicit about as a project-wide pattern.

• A future v0.3 preprint-policy-compliance skill, if it emerges, would consume this note’s framework. The per-journal-policy-axis content here could be extracted into a structured machine-queryable resource as a follow-up; for v0.2 it stays as a knowledge note that the skill reads.

• The fast-moving-landscape framing — “principles in this note; specifics verified at invocation time” — is the right posture for several other areas where training-data knowledge ages faster than the discipline does (the AI-agentic-scientific- writing landscape, the predatory-publishing landscape, OA policy specifics). Worth being explicit that this is a general scriptorium pattern.

Condition that would flip this: if the reviewed-preprint model (PCI, Review Commons, eLife post-2022, F1000Research) becomes the dominant publication mode in any major field, the venue-fit skill’s whole shape would shift — the question becomes “which review community / platform” rather than “which journal”. Worth re-reading this note in 18 months and updating the emphasis accordingly. The 2025–2026 funder policies (the HHMI requirement and the Gates expectation) move part of the life-sciences author population toward that condition from the funder side, without yet making it dominant.

Cross-references

• venue-selection — primary grounding for venue-fit; preprint recommendation is the additional layer this note adds.
• predatory-publishing — same skill applies predatory refusal to preprint servers as well as journals (most preprint servers are not predatory in the usual sense, but the question is worth applying — the OA preprint ecosystem has predatory analogues).
• editorial-decision-making — preprint policies interact with journal triage; some editors look for preprint citation patterns in their desk-rejection heuristics.
• declared-work-scope — the convention. Preprint recommendation operates on declared work; the skill does not preprint anything on behalf of the author.

References

[1] Fraser, N., Brierley, L., Dey, G., Polka, J. K., Pálfy, M., Nanni, F., & Coates, J. A. (2021). The evolving role of preprints in the dissemination of COVID-19 research and their impact on the science communication landscape. PLoS Biology, 19(4), e3000959. DOI: 10.1371/journal.pbio.3000959. PMID: 33798194. (The canonical assessment of how COVID-19 accelerated biomedical preprinting and the implications.)

[2] Abdill, R. J., & Blekhman, R. (2019). Tracking the popularity and outcomes of all bioRxiv preprints. eLife, 8, e45133. DOI: 10.7554/eLife.45133. PMID: 31017570. (Empirical analysis of bioRxiv preprint trajectories and journal publication outcomes.)

[3] Sever, R. (2026). The story behind the science: preprints of pandemic potential — how bioRxiv and medRxiv brought preprints to the life sciences. mBio, 17(2), e02989-25. DOI: 10.1128/mbio.02989-25. (Replaces “Polka et al. 2024, Annual Review of Biomedical Data Science”, which could not be found in Crossref or PubMed and appears not to exist. A first-hand account of how the two servers brought preprints to the life sciences, by their co-founder; the author’s affiliations are openRxiv and Cold Spring Harbor Laboratory.)

[4] Rawlinson, C., & Bloom, T. (2019). New preprint server for medical research. BMJ, 365, l2301. DOI: 10.1136/bmj.l2301. PMID: 31167753. (The BMJ editorial announcing medRxiv with Yale and CSHL. Replaces “Maggio et al. 2020, Medical Education 54(7)”, whose DOI 10.1111/medu.14181 and PMID 32249938 resolve to unrelated papers. A 2018 Maggio, Artino, Picho & Driessen paper with a similar title exists — “Are you sure you want to do that? Fostering the responsible conduct of medical education research”, Academic Medicine 93(4), 544–549, PMID 28678102 — but it is about medical education research, not preprints.)

[5] arXiv. (Ongoing.) About arXiv. https://arxiv.org/about. (Primary source for arXiv’s history, moderation, governance and size: “more than three million scholarly articles in eight subject areas”; independent nonprofit since 2026.)

[6] openRxiv. (Ongoing.) About medRxiv. https://www.medrxiv.org/content/about-medrxiv. (medRxiv’s founders, operator and screening criteria, including screening for material that might pose a health risk. URL updated; the previous /about path returns 404.)

[7] Brierley, L., Nanni, F., Polka, J. K., Dey, G., Pálfy, M., Fraser, N., & Coates, J. A. (2022). Tracking changes between preprint posting and journal publication during a pandemic. PLoS Biology, 20(2), e3001285. DOI: 10.1371/journal.pbio.3001285. PMID: 35104285. (The preprint-to-published-version drift during COVID-19 — concrete evidence of preprint version-control risks. PMID corrected from 35104289, which is a different paper.)

[8] Elsevier. (2016, 17 May). Elsevier acquires the Social Science Research Network (SSRN), the leading social science and humanities repository and online community. Press release. https://www.prnewswire.com/news-releases/elsevier-acquires-the-social-science-research-network-ssrn-the-leading-social-science-and-humanities-repository-and-online-community-579752311.html. (Replaces an unverifiable blog-post pointer. Primary record of the acquisition. The governance controversy that followed is not sourced here.)

[9] Peer Community In (PCI). (Ongoing.) https://peercommunityin.org. (Community-run discipline-specific preprint recommendation system, created by INRAE researchers Denis Bourguet, Benoit Facon and Thomas Guillemaud, the authors of [10]. Corrected in the 2026-10 refresh: the earlier entry named a non-existent co-founder and attributed a “founding paper” to Casadevall & Fang 2014.)

[10] Guillemaud, T., Facon, B., & Bourguet, D. (2019). Peer Community In: a free process for the recommendation of unpublished scientific papers based on peer review. In ELPUB 2019 23rd International Conference on Electronic Publishing. DOI: 10.4000/proceedings.elpub.2019.23. (Formal account of PCI’s design by its founders; replaces the documentation-site placeholder.)

[11] Review Commons. (Ongoing.) https://www.reviewcommons.org. (Journal-independent refereed-preprint platform for the life sciences, developed with ASAPbio and implemented at EMBO; launched in 2019. The journal-independent peer-review model.)

[12] F1000Research. (Ongoing.) https://f1000research.com. (Post-publication peer-review platform operated by Taylor & Francis since 2020.) Founding paper: Tracz, V., & Lawrence, R. (2016). Towards an open science publishing platform. F1000Research, 5, 130. DOI: 10.12688/f1000research.7968.1.

[13] Eisen, M. B., Akhmanova, A., Behrens, T. E., Diedrichsen, J., Harper, D. M., Iordanova, M. D., Weigel, D., & Zaidi, M. (2022). Peer review without gatekeeping. eLife, 11, e83889. DOI: 10.7554/eLife.83889. PMID: 36263932. (The October 2022 eLife editorial announcing the end of accept/reject decisions after peer review.)

[14] SHERPA/RoMEO (now Open Policy Finder, operated by JISC). (Ongoing.) https://openpolicyfinder.jisc.ac.uk/. (Publisher preprint and self-archiving policy database. The authoritative reference for journal-by-journal preprint policy.)

[15] cOAlition S. (Ongoing.) Plan S Principles and Implementation. https://www.coalition-s.org. (The cOAlition S funder consortium open-access requirements that drive preprint adoption among funded researchers in Europe.)

[16] National Institutes of Health (NIH). (2024; effective 1 July 2025). 2024 NIH Public Access Policy. https://publicaccess.nih.gov. Policy notice NOT-OD-25-047 (original effective date 31 December 2025; requirements): https://grants.nih.gov/grants/guide/notice-files/NOT-OD-25-047.html. See also NOT-OD-25-101, Notice of updated effective date for the 2024 NIH Public Access Policy: https://grants.nih.gov/grants/guide/notice-files/NOT-OD-25-101.html.

[17] Howard Hughes Medical Institute (HHMI). (Ongoing.) Research policies: publishing and sharing. https://www.hhmi.org/about/policies/publishing-sharing. (Index page for HHMI’s Open Access to Publications Policy, effective through 31 December 2025, and its Immediate Access to Research Policy, effective 1 January 2026 [26].)

[18] Fu, D. Y., & Hughey, J. J. (2019). Releasing a preprint is associated with more attention and citations for the peer- reviewed article. eLife, 8, e52646. DOI: 10.7554/eLife.52646. PMID: 31808742. (Empirical evidence on preprint-citation relationship.)

[19] Fraser, N., Momeni, F., Mayr, P., & Peters, I. (2020). The relationship between bioRxiv preprints, citations and altmetrics. Quantitative Science Studies, 1(2), 618-638. DOI: 10.1162/qss_a_00043. (Finds higher citation and altmetric counts for bioRxiv-deposited articles, and says further research is needed to establish causality. Replaces “Wang et al. 2024, JMLA 112(1)”, whose DOI is not registered and whose PMID 38486838 resolves to an unrelated retraction notice.)

[20] arXiv. (2026, 1 October). Fair moderation, equitable access, and AI: arXiv’s updated rate limit policy. arXiv blog. https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/.

[21] arXiv. (2025, 31 October). Attention authors: updated practice for review articles and position papers in arXiv CS category. arXiv blog. https://blog.arxiv.org/2025/10/31/attention-authors-updated-practice-for-review-articles-and-position-papers-in-arxiv-cs-category/.

[22] openRxiv. (2025, 11 March). Introducing openRxiv: a new independent home for preprint sharing in the life sciences. https://openrxiv.org/introducing-openrxiv/. News coverage: Naddaf, M. (2025). Preprint sites bioRxiv and medRxiv launch new era of independence. Nature (news). DOI: 10.1038/d41586-025-00762-4.

[23] eLife. (2024, addendum 2025). Changes to eLife’s indexing status in Web of Science and Scopus. https://elifesciences.org/inside-elife/ae620829/changes-to-elife-s-indexing-status-in-web-of-science-and-scopus.

[24] National Institutes of Health. (2025). NOT-OD-25-138: Request for information on maximizing research funds by limiting allowable publishing costs. https://grants.nih.gov/grants/guide/notice-files/NOT-OD-25-138.html. Status as of March 2026: Wood, E. (2026, 20 March). Guest post — All the seats at the table: a summary and status review of the NIH APC caps proposal. The Scholarly Kitchen. https://scholarlykitchen.sspnet.org/2026/03/20/guest-post-all-the-seats-at-the-table-a-summary-and-status-review-of-the-nih-apc-caps-proposal/. Re-checked 20 October 2026 against NIH’s list of selected policy notices, https://grants.nih.gov/policy-and-compliance/notice-of-policy-changes, which lists no notice finalising publication-cost limits. The list is described as “selected” notices, so absence there is strong but not conclusive evidence that no cap has been finalised.

[25] Wellcome. (Ongoing.) Open access policy. https://wellcome.org/research-funding/guidance/ending-a-grant/open-access-guidance/open-access-policy. (APC rule effective 1 January 2025. The page showed no last-updated date when re-checked in October 2026.)

[26] Howard Hughes Medical Institute. (2025; effective 1 January 2026). Immediate Access to Research Policy. https://hhmicdn.blob.core.windows.net/policies/Immediate-Access-to-Research.pdf. (Issue date 24 September 2025; the PDF shows “Last Updated: 08/05/2026”, i.e. 5 August 2026. Applies to the HHMI roles listed in the policy and to original research articles that are major contributions from their labs, (co-)first, (co-)last or (co-)corresponding author; it does not apply to review articles, perspectives, commentaries and similar.)

[27] Gates Foundation. (Ongoing; refreshed policy effective 1 January 2025). Open Access Policy. https://openaccess.gatesfoundation.org/open-access-policy/. (“It is the foundation’s expectation that Funded Manuscripts shall be shared promptly in the form of a preprint … except where the grantee determines that a preprint is not appropriate due to ethical, safety or other legitimate concerns.” Re-checked 20 October 2026.)