In short. Use distribution with DataDownload, contentUrl and encodingFormat to describe each downloadable dataset representation.
Definition: Dataset markup uses
distributionto describe a downloadable form of a dataset. Inside eachDataDownloadobject,contentUrlgives the download link andencodingFormatstates the format of that distribution.
Key facts
- Google's Dataset documentation describes
distributionas the place to get the data and the format in which it is available. - Google lists
distribution.contentUrlas a required URL for a download and describes it as “The link for the download.” - Google recommends
distributionas aDataDownloadvalue anddistribution.encodingFormatas text or a URL that identifies the file format. - Schema.org defines DataDownload as “All or part of a Dataset in downloadable form.”
- Schema.org describes
encodingFormatas a media type typically expressed with a MIME format, such asapplication/ziporaudio/mpeg.
What does distribution describe?
The dataset page usually describes the collection. A distribution value describes one downloadable form of that collection, including where to obtain it and its format. Google says that a dataset can have several distribution values: a CSV version can use one URL while an Excel version uses another.
Schema.org's DataDownload type represents all or part of a dataset in downloadable form. This makes each available file or downloadable representation a separate object inside distribution. A dataset with several formats can therefore use an array of DataDownload objects, with each object carrying its own link and format.
Where do contentUrl and encodingFormat go?
Place both properties inside the DataDownload object. Google's documentation lists distribution.contentUrl as required for the download description. It lists distribution.encodingFormat as recommended and accepts either text or a URL. The text can be a short format name such as CSV; a MIME value such as text/csv gives a more specific media type when that value is known.
Google's example shows two separate downloads under one distribution array:
"distribution":[
{
"@type":"DataDownload",
"encodingFormat":"CSV",
"contentUrl":"https://www.ncdc.noaa.gov/stormevents/ftp.jsp"
},
{
"@type":"DataDownload",
"encodingFormat":"XML",
"contentUrl":"https://gis.ncdc.noaa.gov/all-records/catalog/search/resource/details.page?id=gov.noaa.ncdc:C00510"
}
]
The same documentation describes contentUrl as the link for the download. Schema.org's contentUrl property describes the actual bytes of a media object, such as an image or video file. For dataset markup, the URL should correspond to the downloadable representation described by that DataDownload object.
How should the format value be written?
Schema.org's encodingFormat definition says the value is typically a MIME format and gives application/zip and audio/mpeg as examples. A MIME value is useful when the representation has a registered media type. The same definition permits a URL for an unregistered or niche encoding or file format, using the most appropriate explanatory page.
Keep the link and format paired in the same object. If CSV and XML are available at different addresses, give each address its own DataDownload object. This follows Google's example, where each distribution has one encodingFormat and one contentUrl.
Related terms
FAQ
Can contentUrl point to a dataset landing page?
Google describes contentUrl as the link for the download, while url identifies a page describing the dataset. Use contentUrl for the downloadable representation and keep the descriptive page URL on the dataset object.
Is encodingFormat required?
Google lists distribution.contentUrl under required properties and distribution.encodingFormat under recommended properties. A format value remains useful because it identifies what the linked distribution contains.
Can one dataset have several download links?
Yes. Google documents several distribution values for cases such as separate CSV and Excel versions. Each DataDownload object can carry its corresponding contentUrl and encodingFormat.
Sources
- Google Search Central, "Dataset (Dataset, DataCatalog, DataDownload) structured data", 8 September 2026
- Schema.org, "DataDownload", accessed 17 October 2026
- Schema.org, "encodingFormat", accessed 17 October 2026
Check how a page reads to answer engines. Open the GEO score analyzer