Media Literacy, Verification and Disinformation

Apply the five-question verification test consistently.

Trace an online claim to its earliest identifiable source and distinguish the source from later reposting.

Evaluate images, videos, translations, statistics and screenshots for context and authenticity.

Use Academy claim labels and execute a correction workflow.

Pass the fact-check assessment at 85% or higher.

Review the Source Hierarchy, Five-Question Verification Test and Claim Labels.

Bring one social-media claim that you are unsure about; do not bring graphic content without trainer approval.

Install or bookmark the verification tools approved by the Academy for the cohort.

Verification begins by narrowing the claim “This video proves what happened” is not a testable claim. Verification starts by rewriting the claim into a sentence that can be checked: “This 27-second video was recorded at Location X on Date Y and shows Actor Z doing Action Q.” Each component can then be examined independently. Ambassadors should resist the instinct to investigate the emotional conclusion first.

Trace to the source, not to the account you trust A respected journalist, government account, activist or friend may repost something originally produced elsewhere. The repost is not the original evidence. Find the earliest available upload, the issuing institution, the full report or transcript, and the surrounding material. If five websites repeat one unsourced claim, that is one source repeated five times, not five-source corroboration. Source independence test: Ask: “If the first source disappeared, would the second source still possess independent evidence?” If not, you probably have repetition, not corroboration.

Images and video: real content can still mislead

Check whether the material appeared online before the claimed event. Old footage is frequently recycled during war or crisis.

Look for landmarks, weather, language, shadows, road signs, uniforms, license plates and other location/time clues, but do not overstate amateur geolocation.

Find longer versions of short clips. A cut beginning two seconds later can reverse the apparent sequence.

Do not assume a caption was written by the original recorder.

Synthetic or altered media should be treated as unverified until authenticated. AI-generated images are never documentary evidence.

Graphic material requires necessity, dignity and safeguarding review. Do not distribute traumatic imagery merely to “prove” seriousness. 4. Screenshots are clues, not self-authenticating records A screenshot can be cropped, edited, mistranslated or stripped of date and URL. Where possible, preserve the original link, capture the account name, timestamp and surrounding thread, and archive the underlying page. If the original is no longer available, state that limitation rather than treating the screenshot as equal to the original record. 5. Translation is a verification problem A translated quote may change tense, subject, idiom, intensity or legal meaning. For sensitive statements, retain the original-language text and identify the translator or method. Machine translation may be a useful first pass but should not be the final authority for a disputed phrase. Where meaning is material, use a competent human reviewer and document disagreement. 6. Numbers require definitions, not only citations Statistics can be numerically accurate and rhetorically misleading. Before repeating a number, ask: What period does it cover? Who is counted? What is the denominator? Is the figure provisional? Was the methodology changed? Is it a cumulative total or a current count? Does the source distinguish civilians and combatants, allegations and confirmed cases, people and incidents? Casualty and hostage figures in fast-moving conflicts are especially sensitive. Different institutions may publish different categories and later revisions. The ambassador should date the number, name the source, state what the source says it measures, and avoid implying a level of independent verification that does not exist. 7. Reverse reasoning: what would change your mind? Before concluding that a claim is true, write one piece of evidence that would cause you to downgrade it. This simple step exposes motivated reasoning. An advocate who cannot imagine disconfirming evidence is no longer verifying; the advocate is defending a predetermined conclusion. 8. Fact-check workflow 1. Capture the exact claim without repeating inflammatory wording unnecessarily. 2. Identify the original source and preserve it with date/time/context. 3. Classify the source tier and check authenticity/currentness. 4. Search for independent corroboration and contradiction. 5. Check image/video context, translations and numerical definitions as applicable. 6. Apply a claim label: Verified, Supported, Disputed, Unverified, False/Materially Misleading, or Opinion/Interpretation. 7. Draft the minimum responsible conclusion and cite the strongest sources. 8. If you later discover a material error, stop distribution and correct proportionally to the reach. 9. Fictional fact-check lab Sample claim Evidence supplied Expected action / label A: “A major newspaper has confirmed Event X.” Ten social accounts link to the same anonymous screenshot supposedly from the newspaper; no article can be located. Unverified. Locate the original article or obtain independent evidence; do not count reposts as corroboration. Sample claim Evidence supplied Expected action / label B: “This video was recorded today in City Y.” The same video appears in a two-year-old archive from another country. False / materially misleading in present context. The video may be real but the date/location claim is false. C: “Court Z ruled that Person A committed Crime B.” A court judgment is available, but it concerns a procedural motion and contains no such finding. False / materially misleading. Read the operative ruling; do not turn allegations into adjudicated facts. D: “Government data show a 30% increase.” Official release states 30% increase compared with the previous month, but the graphic says “30% increase this year.” Materially misleading. Correct the comparison period. E: “Organization Q says 1,000 incidents occurred.” Organization Q’s methodology is published; a second independent database records a similar trend but different count. Supported for the trend; attribute the exact count to Q and explain methodological differences. 10. Correction protocol 1. Stop further distribution of the incorrect material. 2. Determine exactly what was wrong and why. 3. Correct the original post where technically possible. Do not silently delete a material error unless safety requires removal. 4. Issue a clear correction proportional to the reach of the error. 5. Notify partners who received the incorrect material. 6. Log the root cause and update the fact sheet, workflow or training if needed.

Why are five websites repeating the same original source not necessarily five independent sources?

What is context collapse?

What information should accompany a screenshot where possible?

Why must a casualty statistic be dated and defined?

What does “reverse reasoning” ask you to do?

Can AI-generated imagery be used as documentary evidence?

What should happen after a material factual error is discovered?

What is the minimum Week 5 assessment score for certification eligibility?

They may all depend on one upstream claim and possess no independent evidence.

Real content is presented with false or misleading date, location, sequence or implication.

Original URL, account/source, timestamp/date, surrounding context and archive reference if possible.

Because the figure may change, have a specific category/methodology, and be misread if the time period or population counted is omitted.

Identify evidence that would cause you to change or downgrade your conclusion, guarding against confirmation bias.

No. Synthetic media may illustrate only if clearly labeled and approved, but it is not evidence of a real event.

Stop distribution, identify the error, correct visibly/proportionally, notify affected partners as needed, and log the cause. 8. 85%.

Trace a claim to an original source.

Distinguish repetition from independent corroboration.

Explain why a screenshot or short clip is incomplete evidence.

Apply claim labels without treating “unverified” as “false.”

Write a visible correction without defensive language.

Searching only for evidence that confirms a preferred claim.

Treating official statements as automatically independently verified facts.

Calling a claim false merely because it is not yet verified.

Publishing sensitive breaking claims before source work is complete.

Reuters, Journalistic Standards / Handbook. handbook.reuters.com/

UNESCO Media and Information Literacy resources, including verification and misinformation modules. www.unesco.org/mil4teachers/en/module4/unit5

ReJEWenate Manual v1.0: Source Hierarchy, Five-Question Verification Test, Claim Labels, Correction Protocol and AI/Synthetic Media rules.