Editorial standards

What we choose

A useful thread is central or memorable, can be checked or meaningfully explained, invites curiosity and can be represented fairly. We score those four qualities from zero to two and aim for at least six out of eight for new reads. Existing reads are being assessed against this standard.

Sources first

We prefer original records, research, official data and museum or archive collections. Academic syntheses and reference works provide context. Journalism and organisational pages are used when appropriate to the claim, with their limitations considered. A source link is not a guarantee that a conclusion is correct.

Automated editorial checks

Agents locate the moment, restate the point, inspect sources, check names and numbers, seek contrary evidence and make a second pass for fairness and clarity. Code checks structure, timestamps and links. New publication records require evidence of those passes. We do not describe agent review as human review.

What status labels mean

• Supported: the cited evidence supports the specific point.
• Context: evidence adds qualifications or background.
• Disputed: cited evidence contradicts the specific point.
• Uncheckable: available evidence cannot establish it.
• Background: an explanation rather than a verdict on a claim.

Older reads without a completed status assessment carry no status badge.

Respect for creators

We link to the original episode and use short, attributed quotations only where needed. This release publishes no full transcripts or transcript search files. Search covers PodCompanions’ own reads. See the creator policy.

Corrections and sensitive subjects

We aim for neutral, properly attributed wording. Health, political and contested claims and statements about living people need particular care and independent evidence. Material corrections are dated on the affected read and recorded in the corrections log.