Publishing policy
This site exists to put work in public while it is still in progress. At some venues that is encouraged; at others a public manuscript ends the submission, permanently. The rule is written down rather than decided case by case, and the build refuses to publish an entry that exceeds what its venue allows.
Policies surveyed 6 August 2026. Publisher policies change; rows are re-checked rather than assumed.
The principle
Metadata public, files gated.
Titles, claims, statuses and identifiers are almost never a problem. Full manuscripts are what trigger rules. Splitting there costs almost nothing in discoverability and removes most of the risk.
And the corollary
Decide the venue before posting anything. Nearly every rule in this area is a one-way door: a preprint cannot be un-posted, one server will not withdraw a paper at an editor’s request, another overwrites previous versions irreversibly, and one target journal refuses any manuscript that has been posted anywhere at all.
Post freely
Preprint and identifier, linked from the entry.
- Journal of Integer SequencesPreprint and identifierunknown reviewread at source 2026-08-06
States three separate times that preprints are fine, including that authors may post to arXiv or any other preprint archive at any time.
Preprint-friendly, so the manuscript may live on a server and be linked here. The gate is the AI policy, not the posting rule.
One-way door. An undisclosed or under-disclosed AI contribution can end the relationship with the journal permanently. Settle the disclosure wording before the first submission, not after.
AI policy. The binding constraint. Language models may not write the text of the paper. Model-suggested proof techniques are permitted only with explicit credit to the model inside the text. Undisclosed use means immediate rejection and can mean a lifetime ban. The journal reserves the right to run detection tools.
- Transactions on Machine Learning ResearchPreprint and identifierdouble anon reviewread at source 2026-08-06
Preprints permitted at any time; submissions are public but anonymised.
- ACL Rolling ReviewPreprint and identifierdouble anon reviewread at source 2026-08-06
Since 15 February 2024 there is no anonymity period and no limitation on posting or discussing non-anonymous preprints during review.
One-way door. An optional binding "no non-anonymous preprint" checkbox at submission converts a legal public preprint into a desk rejection. Anything at any stage of this review also bars the same work from COLM.
- COLMPreprint and identifierdouble anon reviewread at source 2026-08-06
Non-anonymous preprints on servers, social media and personal sites are permitted; reviewers are instructed not to search for them, and encountering one is not a conflict.
- IEEE SoftwareIEEEPreprint and identifiersingle anon reviewread at source 2026-08-06
Preprints on a personal site, an employer server, or an approved server are normal. A posted preprint must be updated after publication rather than removed.
The published version of record may never be self-archived; the preprint may.
- Transport FindingsPreprint and identifiersingle anon reviewread at source 2026-08-06
Working papers and technical reports on personal or institutional sites are explicitly fine.
- International Journal of Medical InformaticsElsevierPreprint and identifiersingle anon reviewread at source 2026-08-06
- Health Affairs ScholarOxford University PressPreprint and identifiersingle anon reviewread at source 2026-08-06
- JAMIAOxford University PressPreprint and identifierunknown reviewPARTIALLY VERIFIED
Explicitly includes posting on the author’s own personal website.
- Research Notes of the AASFull manuscripteditorial reviewread at source 2026-08-06
Not peer reviewed; editor-moderated, and welcomes null results and work in progress.
Astronomy and astrophysics only, which limits what can actually go here.
Post, then freeze
Post, supply the identifier at submission, then do not touch the deposited copy until a decision.
- Crime, Media, CultureSagePreprint and identifierdouble anon reviewread at source 2026-08-06
Preprints accepted; supply the preprint identifier at submission and link forward to the published version afterwards.
The publisher states plainly that posting a preprint may let reviewers discover the author, and that this does not prevent consideration. The risk is real and is the author’s to accept.
One-way door. Once posted it cannot be withdrawn to restore anonymity.
- Computational Communication ResearchPreprint and identifierdouble anon reviewread at source 2026-08-06
Preprints permitted at any stage; specific servers are named in the guidance.
- Journal of Official StatisticsSagePreprint and identifierdouble anon reviewread at source 2026-08-06
Disclose the preprint and its identifier at submission, and do not update it during review.
- Health Services ResearchWileyPreprint and identifierdouble anon reviewread at source 2026-08-06
Named servers are acceptable; no changes to the posted version while it is under evaluation.
- Economic Development QuarterlySagePreprint and identifierdouble anon reviewread at source 2026-08-06
Supply the identifier at submission; no changes during evaluation.
Post nothing
Title and status only. The venue is not named publicly while the work is out.
- Television & New MediaSageTitle and status onlydouble anon reviewread at source 2026-08-06
The journal does not accept manuscripts that have been posted on preprint servers. This is a journal-level override of an otherwise permissive publisher policy.
A prohibition that cannot be reversed. Whether a manuscript on the author’s own site counts is genuinely ambiguous, so it is treated as though it does.
One-way door. Posting anywhere, once, permanently disqualifies the manuscript here.
- International Journal of Research in MarketingElsevierTitle and status onlydouble anon reviewPARTIALLY VERIFIED
The publisher’s general policy is permissive but carves out society-owned and double-anonymised titles, telling authors to check the journal. This journal is both, and its own guidance is silent.
Silence is not permission. Ask the editorial office before posting anything.
- Managerial and Decision EconomicsWileyTitle and status onlydouble anon reviewPARTIALLY VERIFIED
Publisher default is permissive; no journal-level statement was found.
- CityscapeUS Department of Housing and Urban DevelopmentTitle and status onlydouble anon reviewread at source 2026-08-06
No preprint policy is stated anywhere, and the journal assigns no identifiers.
Policy silence at a double-anonymous venue is treated as prohibition until asked.
AI policy. Authors must disclose whether AI was used and specifically how, in the acknowledgements.
Where to deposit
Work with no journal yet can still be permanently citable. A dated deposit establishes priority in a way a web page cannot. What differs between services is the gate, the versioning behaviour, and whether anyone will find it.
| Service | Gate | Versioning | Watch out for |
|---|---|---|---|
| Zenodonot indexed for citations | None. Identifier issued in seconds. | Every version gets its own identifier, plus a stable concept identifier that always resolves to the latest. The concept identifier is a separate number, not derivable from the version one. | Not indexed by the main scholarly search engine, so a deposit here accrues no citation counts on its own. Records from unaffiliated depositors rank lower. Generative-AI disclosure is required of depositors. |
| arXiv | Endorsement. Since 2026-01-21 automatic endorsement requires both an institutional address and prior accepted authorship in the same domain, so an unaffiliated first-time author needs a personal endorsement from an established author in that domain. Mathematics is a single domain; its subject classes do not endorse separately. | One identifier for the work, updated in place. No per-version identifier. | - |
| SSRN | Affiliation field accepts "independent researcher". Days to post. | Destructive. A revision overwrites the previous version with no history and no way back, which collides directly with the venues that require a preprint to stay frozen during review. | - |
| SocArXiv | Researcher identifier required; moderated before posting. | New versions get new identifiers. | Social sciences only, with no mathematics branch, and it will not withdraw a posted paper at a journal editor’s request. |
| Research Square | None, but only research articles and systematic reviews are eligible. | Versioned, with a cap on revisions. | Theory papers are explicitly ineligible, which rules out most of the mathematical work here. |
| TechRxivtransitioning | n/a | n/a | No deposits recorded since March 2026; the platform appears to be mid-transition. |
| OSF Preprints (generalist)closed | n/a | n/a | Closed indefinitely since late 2025, attributed to generative-AI submission volume. |
Anonymous review
Preprints and anonymous review are compatible, not invisible. The publisher of several of these journals states plainly that posting may let reviewers discover the author, that this will not usually prevent consideration, and that the author should be comfortable with it beforehand. That is the honest framing, and it puts the choice on the author rather than pretending anonymity survives.
A public analysis repository is a separate exposure from the manuscript. A reviewer who searches a distinctive dataset name or a phrase from the methods can arrive at a repository under a real name, with a commit history and an address in the log; nothing from the paper needs to be in it. Where data must be shared during anonymous review, it is shared through an anonymised mirror or a private link, with file metadata scrubbed.
Identifiers without a journal
A deposit can be made with its files embargoed, producing a live, resolving, timestamped identifier with public metadata and no public full text. For work aimed at a venue in the third bucket that is the right instrument: the work is dated and citable, and nothing a reviewer could match against the manuscript is public.
Where an entry has no identifier at all, its citation block says so and offers the URL and a revision history instead, so a dated citation still resolves to a known state of the page.