Start with the actual assertion
Keep the post’s words and media together. For a comment, preserve the parent post and distinguish the commenter’s claim from the post’s. If the capture is incomplete or points to another target, stop and repair it before research. A correct answer about the wrong post is still a failed check.
Find the record behind the link
Look for the original document, relevant passage, date, population and method. Explain what it supports and what it does not. Trace syndicated stories to their common origin: five copies of one report are not five confirmations.
Try to disprove the attractive answer
Seek the strongest material counterevidence. Compare the same units, place and period. Do not manufacture a midpoint when the evidence favors one conclusion. A small average cannot erase an unusually large local burden.
Describe the framing, not the person
Name a visible cue and explain possible readings. A dramatic image might emphasize danger; a small denominator might reassure. Neither proves the writer’s intent, political allegiance or psychological bias. Missing context remains missing.
Keep the update attached
State what evidence would materially change the finding. Distinguish an original error from new observations, changed circumstances and a forecast reaching its horizon. Reopening a saved report does not search again.
Keep people in the forecast
Show the decisions people actually made and the choices they had. Adaptation can change an outcome; it is not a guaranteed happy ending. Later benefits do not erase earlier harm.
Two historical lanes, different gates
Forecast Files examine a dated warning, genuine hazard, outcome and documented response. The Alarm Archive additionally needs a specific overreach and an attributable response cost. A scary cartoon or a disaster alone does not qualify.
AI assistance is not editorial approval
Automated checks can make mistakes. This redesign’s article copy awaits exact-version human approval. A working source link is not proof that its passage supports the conclusion. The launch still needs an accountable editorial owner and high-stakes escalation rules.