Freeze the claim
Copy the exact wording, date, author and URL before the wording changes.
Claim→Origin→Evidence→Alternatives→Assessment
Do not jump from the first box to the last.
The purpose of research is not to make your first interpretation harder to attack. It is to discover whether it deserves to survive.
The research workflow
Eight moves from discovery to revision.
Use the full workflow for consequential claims. For low-stakes claims, use it as a checklist and stop when further research would not change a decision.
Trace it backwards
Follow citations, screenshots and quotations until you reach the earliest accessible source.
Restore context
Read before and after the excerpt. Check date, location, speaker, audience and whether the material was edited.
Classify the evidence
Separate documents, direct records and measured data from testimony, analysis and speculation.
Test the mechanism
For causal or attribution claims, identify decisions, communications, resources, actions and measurable effects.
Search against yourself
Look for the strongest counterevidence and the best version of a competing explanation.
Compare and calibrate
Ask which explanation accounts for the evidence with the fewest unsupported assumptions.
Preserve revision
Record the date, sources, uncertainty, next check and what would change your view.
Choose the burden of proof
Not every claim asks the same question.
A screenshot may establish that words appeared on a page. It cannot, by itself, establish why they appeared, who authorised them or what will happen next.
Did an event, statement or record exist?
Contemporaneous records, authenticated material and independent corroboration.
Did X produce Y?
Sequence, mechanism, comparison, dose or variation, and exclusion of credible alternatives.
Who was responsible?
Actor-specific access, behaviour, communication, operational links and chain of command.
Will an outcome occur?
A time-bounded forecast, base rate, explicit assumptions and a measurable resolution rule.
What should be done?
Separate factual premises from values, priorities, trade-offs and acceptable risk.
Verification routines
Six checks that prevent avoidable errors.
Tools can reveal provenance, context and manipulation. They cannot replace judgement about what the material actually establishes.
01Find the origin+
Search an exact phrase in quotation marks, remove later date ranges and inspect citations backwards.
The goal is provenance—not the most popular result.02Check source independence+
Draw one line from every article to the evidence it relies on. Count origins, not outlets.
Syndication and circular citation imitate confirmation.03Verify an image+
Use reverse-image search, compare earlier appearances and inspect landmarks, weather, shadows and signage.
Metadata can help, but missing or editable metadata is not a verdict.04Verify a video+
Find the earliest upload, inspect key frames, listen for cuts and compare the claimed time and place.
A real video may be old, cropped, mirrored or described falsely.05Recover a changed page+
Check web archives and cached copies; preserve the exact archived URL and capture date.
An archived page shows what was available—not automatically why it changed.06Read data honestly+
Find the denominator, time range, comparison group, missing values and whether the chart axis distorts scale.
A correct number can still answer a different question from the one being claimed.Private research workbench
Build a conclusion you can inspect later.
Your entries remain in this browser. Nothing is submitted to the site.
0/7research fields completed
Saved on this device
Know when to stop
More searching is not always more rigorous.
Stop when additional information is unlikely to change your decision, when the claim cannot be made testable, or when the conversation has shifted from evidence to identity defence.
A responsible outcome can be: “I do not know, and the available evidence does not justify a stronger conclusion.”