A download link may work today but tell a reader little about who created the file, which analysis it supports or whether it is the version used in a publication. A persistent identifier can provide a stable route to that context when the record is maintained properly.
Decide what the identifier identifies
Before sharing data, define the research object. A raw spectrum collection, a processed peak table and a complete study package may be different objects with different relationships to the publication.
An original example might place raw instrument files and a derived results table in one documented dataset release, with a manifest showing which files support each figure. Alternatively, a repository may organise them as related records.
The important point is that the citation should lead to an intelligible object. Do not cite a broad project page as though it uniquely identifies the exact table used in an analysis.
Choose a suitable repository and follow its deposition requirements. Record the title, creators, release information and the identifier actually assigned; do not invent an identifier while the record is still being prepared.
Inspect the landing page
DataCite recommends that a DOI resolve to a landing page containing identifying metadata and a bibliographic citation, with a way to access the item. Its guidance distinguishes that contextual page from a direct link to a PDF or another file.DataCite — Best Practices for DOI Landing Pages (opens in a new tab)
Open the assigned link as a reader would. Check that the dataset title, creators and description match the intended research object and that the files can be located without private instructions.
| Check | Why it matters |
|---|---|
| Correct record | A valid identifier can still point to the wrong dataset |
| Applicable release | The current files may differ from those analysed |
| File descriptions | Readers need to distinguish raw and derived material |
| Access statement | A public record does not necessarily mean open files |
| Publication connection | The relationship to figures or analyses should be clear |
DataCite allows underlying content to be restricted while requiring a public landing page and open metadata. Therefore, a working DOI alone does not establish that every reader can download the data.DataCite — Best Practices for DOI Landing Pages (opens in a new tab)
Connect the record to the analysis
Place the dataset citation where a reader can understand its role, such as a data-availability statement or the relevant methods section. Explain whether it contains original measurements, processed outputs or supporting documentation.
If an analysis used a particular release, retain the release information in the project record and citation where supported. “Latest version” can change after publication and may not reproduce the historical analysis.
Check that file-level names in the manuscript agree with the deposited package. A persistent record cannot compensate for a figure legend that points to a nonexistent column or an obsolete supplementary file.
Handle later changes visibly
If data are corrected or supplemented, follow the repository’s versioning process and explain the relationship to the earlier release. Keep the publication’s original data connection understandable rather than silently substituting a different analysis package.
DataCite’s landing-page guidance describes retaining an identifying page when content is deliberately made unavailable. Such a page can explain the record’s status even when the original download is no longer offered.DataCite — Best Practices for DOI Landing Pages (opens in a new tab)
When citing someone else’s dataset, inspect its present status and note limitations relevant to your use. If files cannot be accessed, do not imply that you independently checked their contents.
A good dataset citation gives the next researcher a clear starting point and enough context to judge what they have found. The scientific appraisal still depends on the data, methods and evidence themselves.
Sources and further detail
- DataCite — Best Practices for DOI Landing Pages (opens in a new tab)
Official guidance inspected for contextual landing pages, access restrictions and retained records when content is unavailable. The study-file checks are an original application.
Sources checked 20 September 2026. Reporting examples are illustrative unless explicitly identified as supplied-document transcriptions. This article has not undergone independent scientific peer review.