The public record
Corrections
Last substantive update:
Substrate Wire is AI-assembled from linked sources. No human editor approves each item, and automated checks can miss errors. A source link makes a claim inspectable; it does not guarantee the claim is correct.
Report an error
Email [email protected], the publication's existing public corrections address. Include the page URL, headline or record, the wording you dispute, and a source supporting the correction. If possible, include the date you saw it.
Reports do not automatically change a page. No response time or individual acknowledgement is promised. Avoid sending sensitive personal information.
Review and correction policy
- Check the disputed wording against its linked source and any relevant primary evidence. Distinguish an error in Substrate's summary from a claim accurately attributed to a publisher.
- If a material claim lacks support, remove it from promoted coverage or mark it disputed while checking. Do not turn an uncertain claim into a confirmed one to resolve a disagreement.
- For a confirmed material error, correct or withdraw the item and attach a dated notice identifying the original claim, the change, the reason, and supporting sources. A withdrawal should explain why the claim cannot be supported.
- Preserve prediction history. Record changed evidence or outcomes as amendments rather than silently rewriting the original prediction or counting unresolved entries as correct.
What needs a notice
Corrections to a number, attribution, publication date, category, or evidence status need a visible notice when they materially change the reader's understanding. A spelling or formatting fix that does not change meaning may be made without a notice. A developing story's new information is an update, not automatically a correction.
This page defines the correction policy. It is not a complete historical correction log. Material notices belong with the affected page or record so readers can see the original context.
General enquiries: [email protected]. For selection criteria and uncertainty labels, read How it works.