Data & access

The same evidence, wherever it is read.

A person and a piece of software should be able to read the same finding, trace it to the same evidence, and see the same limits.

Business records, downloads, and an API are not available yet. This page sets out the proposed access commitments.

Make the public page useful on its own.

The essential record should be readable as ordinary web content. It should explain the business’s identity, the scope of each finding, its sources, and what remains unknown. Reading those facts should not require a paid account.

Important qualifications should sit beside the claims they limit. People should not have to open a separate policy page to discover that a claim is old, disputed, or based only on the business’s own statement.

Keep structured data faithful to the page.

Any machine-readable record should come from the same underlying information as the public page. It should not make stronger claims, omit a dispute, or turn an unknown value into a positive or negative answer.

We propose including:

• Stable identifiers for the business, location, and individual claims.
• The scope and text of each finding, with its source and source type.
• When the evidence was checked, the period it covers, and any expiry or review due date.
• Uncertainty, conflicting evidence, and the status of a dispute.
• The method version used and the record’s material revision history.
• A link back to the public record so a reader can inspect the context.

Make changes possible to follow.

We intend to explore versioned exports, documented access, and a correction feed so systems using the data can discover changes. Stable links should make it possible to cite a specific record or revision without losing the current record.

Formats, endpoints, access limits, and availability commitments have not been set. When access is offered, those details should be documented and tested before anyone is asked to depend on them.

Respect the limits of sharing.

A public source is not automatically free to redistribute. Access would depend on source rights, privacy, consent where required, and applicable law. Some records may support a finding without being suitable for republication.

Where evidence cannot be shared, the record should explain that limit. We would publish reuse terms before offering downloads and distinguish public findings from restricted source material.

Test whether the meaning survives.

We propose testing whether software preserves dates, scope, uncertainty, and disputes when retrieving or summarizing records. An answer that drops those qualifications can be misleading even when it copies the main claim correctly.

Publishing clear data cannot guarantee that a search engine will index it or that an AI system will cite it accurately. Our responsibility would be to make the source record accurate, accessible, and easy to check, and to document known problems in how it is reused.

References informing this proposal

The W3C Data on the Web Best Practices informs the proposed approach to provenance, identifiers, versions, and access. Google’s guidance on AI features explains the importance of accessible text and agreement between visible content and structured data. These references do not certify or endorse Standard Local.