Rudra Kasturi Inc.

Article · AEO

Dataset markup: how to make your data easier to find in Google Dataset Search

Illustrated avatar of Sofia Alvarez

By Sofia Alvarez · Reviewed by Editorial Desk
November 12, 2026 · 2 min read

SUMMARY

Google's dataset documentation (last updated 8 September 2026) says datasets are easier to find in Dataset Search when you supply a name, description, creator and distribution formats as structured data. Name and description are required, the description must be 50 to 5000 characters, and creator, license and distribution are recommended. The steps are ours.

Research and government sites publish data that other people want to cite and reuse. Google’s dataset documentation (last updated 8 September 2026 as we read it) describes how to make those datasets discoverable.

Best for: research institutions, public bodies, data journalism teams and companies that publish open data.

Dataset markup: name it clearly, state creator and license, link the download, help Google find the pages

What Google says

  • Datasets are easier to find in Dataset Search when you supply information such as name, description, creator and distribution formats as structured data. The aim is to improve discovery of datasets from fields such as life sciences and social sciences.
  • Google understands schema.org Dataset markup, with JSON-LD in schema.org syntax preferred. It also has experimental support for W3C CSVW.
  • Required: name (a descriptive name; use unique names for distinct datasets whenever possible) and description (a short summary of 50 to 5000 characters, which may include Markdown, with absolute URLs for embedded images).
  • Recommended properties include creator, license, identifier, isAccessibleForFree, keywords, sameAs, temporalCoverage, spatialCoverage, variableMeasured, url and distribution. For distribution (DataDownload), contentUrl is required and the file format is described.
  • Use a sitemap, and sameAs markup, to document how dataset descriptions are published across your site, and follow Google’s source and provenance practices.
  • To keep a dataset out of Dataset Search, use the robots meta tag. Changes may take days or weeks to appear.
  • Google’s example describes a collection of files that together form a meaningful dataset, and uses sub-datasets for parts.

What to do: step by step

StepOwnerTimeHow to checkDone when
1. Inventory datasetsData steward2 hoursOne row per dataset with a landing page URLA register exists
2. Write unique namesData steward1 hourSort names and look for duplicatesNo two datasets share a name
3. Write the descriptionsResearcher3 hours50 to 5000 characters eachAll descriptions meet the length
4. Add creator and licenseLegal, data steward1 hourA license URL per datasetNo missing license
5. Add distributionDeveloper2 hoursEach download has a contentUrl and formatLinks work
6. Add coverage and variablesResearcher2 hoursTime span, place and measured variablesOptional fields filled where known
7. Publish a sitemap and validateDeveloper, SEO1 hourSitemap lists landing pages; Rich Results Test passesSamples validate

Common mistakes

  • Giving different datasets the same name.
  • A description that is a title in disguise, under 50 characters.
  • Linking to a page rather than the file in contentUrl.
  • Leaving out the license, which leaves reuse unclear.
  • Marking up pages that are not about a dataset.

Related: the schema decision table, XML sitemaps and organization markup. Dataset Search visibility is not guaranteed.

Sources

More on AEO

See all →