Prepublication accuracy checklist for magazine articles

What to take away

• Check every element readers can interpret as factual.
• Prefer original records and direct knowledge, but test their limits.
• Statistics require definitions, denominators, dates, methods, and uncertainty.
• Fairness includes context, response, proportion, and honest sourcing.
• Recheck the final layout, not only the edited manuscript.

This checklist is a release gate, not a substitute for reporting and editing. Assign each item to a named person, record evidence and decisions, and stop release when a consequential claim lacks support. The level of corroboration should rise with the claim's harm, dispute, novelty, or difficulty to correct.

Claim and source record

• Every checkable sentence, heading, caption, graphic, and callout is marked.
• Each marked claim has a specific source, page, timestamp, or direct observation note.
• Summaries have been traced to underlying documents where practical.
• Source identity, authority, date, motive, and access are understood.
• Anonymous or confidential sourcing follows the publication's approval rules.
• Conflicting accounts are resolved, attributed, narrowed, or disclosed.

The Journalist's Resource's guide to critical thinking and research argues that verifying information and putting it in context has strong parallels with academic method, because both gather data and then test tentative explanations until a defensible one remains. It also notes that knowing the precise meaning of research terms helps a journalist read and evaluate a study rather than repeat its summary. A release gate applies that reasoning to the whole package, not only the body copy.

People, quotations, and scenes

• Names, spellings, pronouns, ages, titles, affiliations, and relevant dates are confirmed.
• Quotations match recordings, notes, documents, or clearly identified recollection.
• Ellipses and cuts do not change meaning or hide context.
• Scene details have a known observer, record, or other stated basis.
• Thoughts and motives are attributed to evidence rather than written as omniscient fact.
• Relevant subjects received a fair opportunity to answer consequential claims.

Numbers and statistics

• Units, currency, time period, population, and geography are explicit.
• Percent is not confused with percentage-point change.
• Totals, subtotals, table labels, and calculations reconcile.
• Denominators and useful baselines are included.
• Survey sample, method, collection date, weighting, and uncertainty are reviewed.
• Causation is not inferred from correlation or sequence alone.

The U.S. Census Bureau's lesson on understanding margin of error explains that a sample estimate's margin of error describes uncertainty from observing a sample rather than an entire population. When comparing estimates, the checker should inspect the stated uncertainty and methodology before declaring a meaningful difference.

Text and package consistency

• Dates, numbers, terms, and names agree throughout the article.
• Headline and deck do not strengthen or simplify the finding beyond evidence.
• Pull quotes preserve exact wording and attribution.
• Captions identify the correct people, place, date, creator, and context.
• Images are not cropped or edited in a misleading way.
• Links open the exact supporting pages and are not mistaken for proof by proximity.

Release and correction readiness

• All queries are resolved in the accepted text.
• Late edits received the required factual and editorial recheck.
• The final proof was compared against the checked manuscript.
• Source packets, permissions, approvals, and decisions are preserved.
• A public correction route and internal owner are known.
• Syndication or downstream teams can receive a correction promptly.

Common questions

Who should sign the checklist?

The people responsible for reporting, editing, checking, visuals, production, and final release should sign their assigned parts.

Is one official source enough?

Sometimes. Test whether the official has direct knowledge, whether the claim is disputed, and what independent evidence exists.

Should low-risk facts be annotated?

Yes, though the depth of checking can be proportionate. Basic errors still damage trust.

What happens when a late headline changes?

Return it to the editor or checker responsible for claim accuracy before release.