Learn

How to cite datasets and software (DOI vs URL)

Dataset and software citations break more often than paper citations because URLs change, versions drift, and repositories move. The goal is to cite something stable and reproducible.

Prefer a DOI when it exists

If the dataset or software release has a DOI (e.g., via DataCite/Zenodo), cite the DOI for stability.

DOIs are designed to remain resolvable even if hosting changes.
Many repositories mint DOIs per release/version—prefer version DOIs for reproducibility.

When a URL is acceptable (and what to include)

If there is no DOI, cite a canonical landing page URL plus version information.

Include version/tag/commit hash for software when possible.
Include access date for URLs that can change.
Prefer the project’s official website or repository release page over a direct file download link.

How to verify you’re citing the right thing

For datasets/software, “exists” is not enough—you want the right version and the right artifact.

Check that the landing page matches the project name and owner/maintainer.
Confirm the version/release date corresponds to what you used.
Prefer stable identifiers (DOI) when available; otherwise add version + access date.

You can run DOI/URL values through Citation Verification as a conservative check.

Common mistakes

Citing a homepage URL without a version when the software changes rapidly.
Citing a direct download link that expires or requires authentication.
Using a DOI for a concept record when a version DOI exists (less reproducible).

Next steps

After you’ve verified identifiers/URLs, generate formatted citations via Citation Generator.