The REF 2029 open access policy in brief
REF 2029 continues to require open access for in-scope journal articles and conference contributions with an ISSN, published between 1 January 2021 and 31 December 2028. Other output types, such as monographs and book chapters, are outside the REF 2029 open access requirement. ([2029.ref.ac.uk][1])
For outputs published from 1 January 2026 onwards, the revised policy shifts the key deposit deadline. Where an output is made compliant through repository deposit, a copy should be deposited within three months of publication, rather than within three months of acceptance as was the case under the earlier REF 2021-style requirement. ([2029.ref.ac.uk][1])
That change may look simple, but operationally it creates a number of challenges. Repository teams still need the accepted manuscript. They still need accurate acceptance and publication dates. They still need to know which version can be shared, which embargo applies, which panel rules may be relevant and whether an exception is justified.
In practice, the compliance workflow remains highly dependent on good communication between authors, CRIS systems, repository systems and open access staff.
The author remains central to compliance
One of the strongest messages from the interview was that the author is still the key person in the process.
The author is typically the person who has the accepted manuscript. The author knows when the article was accepted. The author receives the publisher correspondence. And, in many institutions, the author is the person who must push the publication record from the CRIS system into the repository.
This is important because REF 2029 may put more emphasis on the publication date for post-2026 outputs, but that does not remove the need for early author action. Waiting until the publication date is known can still create risk. If the accepted manuscript has not been captured, if the record is incomplete, or if the publication date is only noticed months later, the institution may already be behind.
The safest operational message to authors is therefore not “you can wait until publication”. It is closer to this: Please deposit the accepted manuscript as soon as you have it. The formal deadline may depend on the policy period and publication date, but early deposit gives the institution the best chance of managing embargoes, licences and compliance correctly.
The REF guidance explicitly allows institutions to continue deposit-on-acceptance workflows where established institutional practice and systems are in place, even though the minimum post-2026 requirement is deposit within three months of publication. ([2029.ref.ac.uk][1])
For repository teams, that is an important distinction. REF 2029 may define the minimum requirement, but institutional workflows can still aim higher.
CRIS systems help with discovery and nudging, but repositories hold the compliance asset
Most UK institutions use a CRIS system such as Symplectic Elements alongside their repository. This combination is powerful.
The CRIS system can detect new publications from external sources such as Scopus. It can notify authors. It can manage open access monitoring fields. It can record exception categories. It can give research offices and REF teams visibility into what has been published and what still needs action.
But the repository remains the place where the actual compliance asset is often managed: the accepted manuscript or version of record, the embargo, the public metadata record and the eventual open access release.
In the interview, the repository manager described material arriving both from Elements and from a dedicated web form. Incoming records are validated. Metadata, deposit dates, acceptance dates, publication dates, embargo periods and exception information are tracked. In their case, part of that compliance checking still happens in an external spreadsheet.
That is not unusual. REF compliance is rarely a single-system problem. It is a process problem that spans systems.
A CRIS can nudge. A spreadsheet can help monitor edge cases. A REF group can decide policy exceptions. But the repository is still where the manuscript, access status and embargo outcome become visible and auditable.
Accepted manuscripts matter more, not less
Rights retention strategies are becoming increasingly important for UK institutions. They can reduce the number of publisher embargo problems and give institutions more confidence that manuscripts can be shared under acceptable terms.
However, a rights retention strategy is only useful if the institution actually has the accepted manuscript.
That was a key point from the interview. Rights retention cannot be meaningfully enforced on a publisher PDF that carries publisher copyright restrictions. The strategy applies to the author’s accepted manuscript, not to any version that the author happens to upload later.
This means repository workflows still need to encourage authors to provide the AAM as early as possible. The REF guidance defines the AAM as the final peer-reviewed text, also known as the final author version or post-print, and states that preprints do not meet the requirement unless they are the AAM. ([2029.ref.ac.uk][1])
For institutions implementing rights retention from 2026 onwards, this creates a practical message: Rights retention reduces embargo complexity only when the institution captures the right version of the work.
DSpace repositories can support that strategy by making version identification explicit in deposit workflows, by storing file-level access conditions, and by managing embargo release in a predictable way.
Gold OA is compliant, but repository deposit is still useful
The REF 2029 policy is clear that where an output is made fully open access on publication, commonly called Gold, Diamond or Platinum OA, and it meets the discovery, access and licensing criteria, no further action is required for REF 2029 open access compliance. ([2029.ref.ac.uk][1])
That matters. Institutions should not create unnecessary administrative work by requiring a duplicate repository deposit in every Gold OA case purely for REF compliance.
At the same time, many repository managers will still prefer to have a repository record, and sometimes a copy, for institutional completeness. There are good reasons for that: preservation, reporting, institutional visibility, internal analytics, researcher profiles and a safer audit trail.
The distinction is important:
Gold OA may be sufficient for REF compliance. Repository capture may still be valuable for institutional management.
A pragmatic workflow could therefore be:
1. Confirm that the publisher version is openly available.
2. Confirm that the licence meets the REF requirements.
3. Record the compliance route in the CRIS or OA monitoring system.
4. Still create or enrich the repository record where this supports institutional policy, reporting or long-term stewardship.
Exceptions need governance, not improvisation
REF 2029 includes a tolerance allowance: for each submission to a unit of assessment, institutions may submit up to 5% non-compliant in-scope outputs, or one non-compliant output, whichever is higher. But the policy also distinguishes this from permitted exceptions. Where a valid policy exception applies, that is treated separately from the tolerance for non-compliant outputs. ([2029.ref.ac.uk][1])
This difference matters operationally.
An exception is not simply a convenient label for “we missed the deadline”. It should be tied to the actual exception categories and supported by a process that can withstand audit. The current REF guidance also notes that the full audit guidance is still being developed, but that audit may focus on whether institutions have effectively managed open access compliance and can provide evidence for their processes. ([2029.ref.ac.uk][1])
The interviewed institution already has a REF strategy group meeting regularly to discuss cases where exceptions may apply. That is exactly the kind of governance structure institutions should consider.
A good exceptions process should answer questions such as:
- Who is allowed to approve an exception?
- Where is the exception recorded?
- Is it recorded in the CRIS, the repository, an external tracker or all three?
- Is the evidence stored?
- Can the institution distinguish between a valid exception and use of the 5% tolerance?
- Can the same decision be explained later during audit?
In the interview, exceptions were managed in an external spreadsheet and also recorded in the OA Monitor area of Elements. That kind of dual recording may feel repetitive, but it reflects the practical need to combine detailed operational tracking with institutional reporting.
DSpace is well suited to the repository role in REF workflows
DSpace repositories are already used by many institutions as institutional repositories. For REF 2029, several DSpace capabilities are particularly relevant.
First, DSpace has mature embargo functionality. Files can be deposited while closed, with metadata discoverable and the full text automatically released when the embargo expires. That aligns well with REF workflows where closed deposit may be compliant as long as the embargo is within the permitted maximum and the file becomes available after expiry.
Second, DSpace supports repository records that are discoverable even before full text is openly available. The REF policy requires outputs to be discoverable by readers and automated tools such as search engines. ([2029.ref.ac.uk][1])
Third, DSpace can support “request a copy” workflows. This is valuable where the manuscript has been deposited but cannot yet be made openly available. It allows the repository to provide a managed access route while respecting embargo restrictions.
Fourth, DSpace can integrate with CRIS systems and institutional workflows. In many institutions, the CRIS is the discovery, author-nudging and monitoring layer, while DSpace is the repository and access-control layer. That division of responsibilities works well when the metadata mappings, deposit workflows and status synchronisation are properly configured.
Finally, DSpace gives institutions control. REF compliance should not depend only on publisher platforms remaining stable, external metadata feeds being complete, or authors remembering details years later. A well-managed institutional repository gives the university its own authoritative record of what was deposited, when, under which conditions and with which access status.
The unresolved question: technical harvesting requirements
One area where the sector still needs clarity is technical metadata and harvesting.
The current REF 2029 open access policy explains the compliance requirements, but it does not yet appear to provide detailed technical guidance on repository harvesting criteria, OAI-PMH requirements or specific metadata formats such as the latest RIOXX profile.
This matters because repository configuration changes take time. If additional technical requirements are published late, institutions and repository providers will have less time to implement, test and validate them.
For DSpace repositories, this may eventually mean reviewing OAI-PMH crosswalks, metadata exposure, licence fields, embargo information, version metadata and compliance indicators. These are usually manageable configuration or development tasks, but they should not be left until the final stages of REF preparation.
The longer the sector waits for detailed technical guidance, the more pragmatic that guidance will need to be. REF 2029 needs to support a broad range of institutions and repository infrastructures. The policy itself recognises the need for flexibility through exceptions, tolerance and audit processes. Technical requirements should follow the same spirit.
Practical recommendations for institutions using DSpace
For institutions preparing their DSpace repository and wider publication workflow for REF 2029, several practical steps are worth considering now.
First, keep encouraging deposit on acceptance. Even if the post-2026 REF deadline is linked to publication, early capture of the AAM remains the safest way to avoid missing manuscripts, unclear dates and preventable embargo problems.
Second, make the accepted manuscript easy to identify. Authors often do not naturally understand the difference between preprint, AAM and publisher PDF. Deposit forms, help text and review workflows should make this distinction explicit.
Third, align CRIS and repository metadata. Acceptance date, publication date, deposit date, version, licence, embargo end date, panel assumptions and exception status should not live in disconnected silos.
Fourth, define an exceptions governance process. REF exceptions should be reviewed consistently, recorded clearly and supported by evidence.
Fifth, review embargo automation. DSpace should be configured so that full text is released automatically and reliably when embargoes expire.
Sixth, prepare for technical metadata changes. Even without final harvesting guidance, institutions can already review the quality of OAI-PMH exposure, licence metadata, open access status fields and version metadata.
Conclusion: the repository is not a REF afterthought
REF 2029 compliance is often discussed in terms of policy, CRIS workflows and institutional reporting. Those are all essential. But the institutional repository remains a central part of the practical compliance infrastructure.
For Green OA, the repository is where the accepted manuscript is stored, embargoed, discovered and eventually opened. For rights retention, it is where the institution captures the version that rights retention can actually apply to. For exceptions and audit, it provides part of the evidence trail. For Gold OA, it may not always be strictly required, but it still contributes to institutional stewardship and visibility.
DSpace repositories are therefore not merely a publication archive in the REF process. They are an operational compliance platform.
As REF 2029 approaches, institutions that invest in strong DSpace workflows, clear CRIS integration, early manuscript capture and disciplined exception governance will be in a much better position to manage the policy confidently, rather than scrambling output by output at the end.
Ready for REF 2029?
Whether you're reviewing repository workflows, improving CRIS integrations, implementing rights retention strategies, or preparing for future REF requirements, now is the time to assess whether your repository infrastructure is fit for purpose.
Atmire helps universities worldwide optimise and extend their DSpace repositories to support open access, research visibility, compliance, and long-term stewardship.
Get in touch with our team below to discuss how DSpace can support your REF 2029 strategy.
[1]: https://2029.ref.ac.uk/guidance/ref-2029-open-access-policy/ "Section 5 – Open Access policy – REF 2029"