The most consequential line in the UK's next research assessment is not about open access at all. It is about a deposit. From 1 January 2026, a journal article or conference proceeding with an ISSN accepted for publication falls under the REF 2029 rule: deposit the accepted manuscript in an institutional or subject repository within three months of acceptance. The earlier demand that the uploaded manuscript be made freely readable on a fixed embargo clock has gone. The update is a deposit rule wearing an open access name.
That distinction will irritate open access advocates and relieve research managers in equal measure. But it is the plain reading of the policy published by Research England, the Scottish Funding Council, the Higher Education Funding Council for Wales, and the Department for the Economy Northern Ireland.
The free-to-read mandate is gone; the acceptance clock is not
REF 2021 operated a two-step rule. Authors had to deposit the accepted manuscript within three months of acceptance. Then the deposited file had to become open access after a maximum embargo period, twelve months for Main Panels A and B and twenty-four months for Main Panels C and D. The REF 2029 update removes those fixed embargo deadlines from the assessment rules. The deposit step remains. The public access step now belongs to funder and publisher policies, not REF compliance.
The target is the author accepted manuscript, often called the AAM: the final peer-reviewed version with all revisions, before the publisher applies typesetting and branding. Deposit is timed from acceptance, not publication. A manuscript accepted on 15 March 2026 must be in the repository by 15 June 2026. It can be closed or embargoed. The repository record still counts for assessment purposes.
What changed and what did not
The official REF 2029 open access policy keeps the scope narrow. It covers journal articles and conference proceedings with an ISSN. Books, book chapters, data sets and other research outputs remain outside this particular deposit requirement. That boundary has not shifted since the 2021 policy. Outputs accepted before 1 January 2026 are outside the new rule and will not suddenly need a retroactive repository record under REF alone.
What changed is the compliance burden. Under REF 2021, institutions and authors chased embargo limits, publisher exceptions and repository technicalities to make files public on time. The 2029 policy gives institutions a quieter task: store the file, record the metadata, and ensure the acceptance date is verifiable.
The repository file does not have to be the publisher version of record. The accepted manuscript is enough. For researchers who have signed copyright to a publisher, that matters because a file can exist inside the institution's system even when the published version sits behind a paywall.
Here's the catch
The catch is that UKRI's separate open access policy has not been relaxed. Since April 2022, peer-reviewed research articles funded by UKRI must be made freely available with a Creative Commons licence immediately on publication, with no embargo. Deposit-only REF compliance will not satisfy that condition. A university can be REF-compliant and still fail the funder's open access audit for the same paper.
That means two systems sit side by side. An unfunded paper may sit in a repository, closed and compliant. A UKRI-funded paper must be open from day one, either through a paid open access route in the journal or by depositing the accepted manuscript with a CC BY licence under a rights retention statement. Research managers are already drafting separate instructions. The separate UKRI open access monographs policy adds another track for long-form outputs.
Photo by Patrick Robert Doyle on Unsplash
The details that trip people up
In practice, the date of acceptance starts the REF clock, not the date of first online publication. Authors who mistake those two dates produce late deposits that become institutional audit findings. Research support offices have spent years fixing exactly this in current research information systems. A manuscript accepted on the final working day before a deadline does not reset because the publisher delays the issue. The acceptance date is the anchor.
Conference proceedings cause confusion because only items with an ISSN are captured, and many accepted papers never receive one. A large academic society's annual proceedings may qualify while a local workshop collection may not. Checking the ISSN before filing the repository record is the reliable test.
The policy also expects the repository record to contain machine-readable metadata with the DOI and journal or series title, including any funder acknowledgement attached to the record. A file dropped into a personal web page or a lab site does not count. It has to live in an institutional repository or a recognised subject repository such as arXiv or PubMed Central.
Exceptions are narrower, not gone
No deposit rule is absolute. The funding bodies have not abolished exceptions, but the 2029 framework removes the old panel-by-panel embargo exceptions. What remains is a practical set: an accepted manuscript may be impossible to deposit when a publisher has supplied no citable version, when a co-author cannot be reached, or when legal restrictions prevent the file from being stored. The rules expect institutions to record the reason rather than silently skip the record.
That record-keeping matters because auditors can accept a documented exception. They cannot accept a missing record with no explanation. The same logic applies to the date of acceptance: institutions need evidence, not assertion.
What research offices should do now
Institutions have a runway until 1 January 2026. The practical work is unglamorous: configure repository workflows and test the metadata template. Ideally the acceptance-email-to-deposit handoff becomes automatic. The less a researcher has to think about the repository, the cleaner the compliance data will be.
Because the REF policy does not override funder rules, the internal guidance should say one thing first: check the funder. UKRI, Wellcome, Cancer Research UK, the National Institute for Health and Care Research and other major UK funders have their own public access conditions. The REF deposit is the floor, not the ceiling. For a UKRI-funded paper, the same acceptance email triggers two actions: deposit the accepted manuscript and check the licence on the publisher's version.
Early-career researchers are least likely to know the difference between a REF deposit and a UKRI open access requirement. A short set of instructions that fits on one screen will outperform a governance document every time.
Hype versus reality
Hype grade: D. Calling this an open access policy flatters it. The rule compels deposit, not access. Headlines that treat it as a green light for free reading will mislead more than they inform.
Reality grade: B. A mandatory repository deposit with a three-month acceptance clock is a useful piece of research infrastructure. It gives institutions a verified bibliographic trail, preserves the accepted manuscript, and creates the only object that can later become open access without another negotiation. That is less than the rhetoric and more than the cynics allow.
The missed opportunity is the lack of any public-access trigger. A deposited file that stays locked until an assessment officer needs it serves administration, not readers. If the goal was open access, the policy is incomplete. If the goal was a reliable audit trail, it works.
Photo by Tim Mossholder on Unsplash
For an early-career researcher building a REF file, the instruction is simpler than the policy debate. When the acceptance email lands, stop. Deposit the accepted manuscript that week. Check the funder's public access condition once. Move on. The academic who does that will not spend next autumn reconstructing dates from a publisher dashboard. They will be writing something new. That is the only version of this reform that matters on a Friday afternoon.