When should a searchable directory use Dataset schema?
Use Dataset markup when the directory exposes a coherent, documented collection of data. A list of marketing cards without data provenance is not a dataset.
Written for teams publishing public directories, research collections, comparison databases, and downloadable records.
Key facts
- Google's Dataset guidance supports discovery of structured data collections in Dataset Search.
- Dataset metadata should describe the collection visible or accessible from the landing page.
- A downloadable file can be represented through DataDownload distributions with formats and URLs.
A useful rule: make each important claim understandable and verifiable without requiring the reader to reconstruct your meaning from the rest of the page.
What separates a dataset from an ordinary directory?
A dataset is a coherent collection of structured observations or records that can be described as a whole. It has a subject, scope, creator, temporal or geographic coverage where relevant, variables or fields, and an access method. A searchable directory may qualify when its entries form such a maintained collection and the landing page documents it. A collection of promotional profile cards does not become a dataset merely because it has filters. Ask whether a researcher could cite the collection, understand how records were gathered, and obtain or query the data under stated terms.
- Define the unit represented by each record.
- Document inclusion rules, coverage, and update cadence.
- Distinguish editorial listings from reusable data.
Which properties make the collection understandable?
Use a clear name and description, then add creator or publisher, license, keywords, spatial and temporal coverage, version, and `dateModified` when they are accurate. A stable landing-page URL should explain the methodology and access. If files are available, model each distribution with DataDownload, `contentUrl`, and `encodingFormat`. Do not copy field names into vague keywords or claim global coverage when the directory contains only selected regions. The structured data should reduce ambiguity about what the collection contains, who maintains it, and what a visitor may do with it.
- Use a specific description with record type and coverage.
- Link a real license rather than writing a vague reuse claim.
- Expose distribution formats exactly as served.
Should every filtered page emit Dataset markup?
Usually no. The canonical dataset landing page should describe the collection, while parameterized filters are views into that collection. Emitting a separate Dataset entity for every filter can create duplicate or misleading identities, especially when filtered URLs are not canonical or indexable. If a stable curated subset has its own methodology, description, and persistent identifier, it may deserve a separate entity. Otherwise, keep one dataset identity and let the interface handle exploration. Align the structured URL with canonical policy so schema does not promote URLs that search systems are asked to consolidate.
- Put the main Dataset entity on one canonical landing page.
- Treat transient search parameters as views, not new datasets.
- Create subset entities only for stable, documented collections.
How should provenance and updates be explained?
State where records come from, how they are cleaned, what may be missing, and when the collection was last refreshed. Link primary source documentation when records aggregate public agencies or research repositories. A visible methodology page is more useful than burying provenance in JSON-LD. Use truthful `datePublished` for the first public release and `dateModified` for material dataset updates, not every build. If coverage changes by region or field, maintain a change log. These details make the collection easier to verify for both human researchers and answer systems that may cite it.
- Publish a concise methodology beside the data.
- Separate source update time from website deployment time.
- Keep a version or change log for material revisions.
What should be tested before publication?
Validate the Dataset JSON-LD, open every license and distribution URL, and verify that files return the advertised formats. Compare record counts and coverage statements with the actual collection. Check canonical tags on the landing page and filtered views. Search for accidental private fields in downloads before exposing them. Then test the directory without JavaScript where practical: the landing page should still explain the dataset and provide crawlable access to core documentation. Success means a user can discover, assess, cite, and access the collection without guessing what the markup meant.
- Verify files, formats, licenses, and access permissions.
- Test canonical behavior for filtered URLs.
- Review privacy and rights before publishing distributions.
Put it to work
Find the highest-impact fix on your site.
Describe the collection, its provenance, license, coverage, and downloadable distributions accurately.
Draft Dataset JSON-LDSources
- 1.Google Search Central: Dataset structured dataChecked 2026-07-26
- 2.Google Search Central: General structured data guidelinesChecked 2026-07-26