
Before you share: give a breaking story five more minutes

The most useful response to an urgent headline is often a short pause. Here is how to make that pause count.
Latest explainers
Browse all →
A reverse image search finds matches—not necessarily the first upload
Search results reveal where an image has appeared, but the oldest indexed match may still be a repost, crop or copy of an earlier source.

A ratio can improve while both underlying numbers get worse
A ratio describes one value relative to another. Its direction can look favourable even when the numerator and denominator both decline.

Never share a one-time code with someone who contacted you
A verification code proves control of your device. A caller asking for it may be trying to complete a sign-in or account reset in your name.

Cleaning up a quotation must not rewrite the speaker’s meaning
Minor edits can improve readability, but changing uncertainty, emphasis or grammar too heavily can turn a real quotation into words the person never gave.

A truncated axis can exaggerate a small difference
Starting a bar chart above zero stretches modest gaps across the plotting area, making similar values appear dramatically different.

A mirrored image can reverse clues without changing the event
A horizontal flip can move signs, traffic and landmarks to the wrong side, creating false location clues even when the underlying scene is genuine.

An annualized rate is not the amount observed so far
Annualizing extends a shorter period as though its pace continued for a full year. It is a projection for comparison, not a count already recorded.

Device encryption protects data best when the device is locked
Encryption can protect files on a lost computer or phone, but an unlocked session may already hold the keys needed to read them.

A source’s silence should not be written as an admission
No response can have many explanations. Reporting should document the contact attempt without claiming that silence proves the allegation.

A dual-axis chart can manufacture a visual relationship
Two vertical scales can be adjusted until unrelated lines appear to rise and fall together. The alignment may come from the axes rather than the data.

A cropped video can hide what happened outside the frame
A clip may show a real moment while excluding the people, actions or surroundings needed to understand why it happened.

A percentage increase can exceed 100 percent
An increase above 100 percent means the added amount is larger than the starting value—not that the result has exceeded some universal maximum.

Read a web address from right to left before you trust it
The registered domain sits immediately before the first slash, and misleading words placed farther left may belong only to a deceptive subdomain.

A headline should distinguish an allegation from an established fact
When a claim remains disputed or unproven, the headline must make that status visible instead of presenting the accusation as a settled conclusion.

Stacked bar charts make middle segments difficult to compare
Only the segments sharing a common baseline are easy to judge. Categories floating in the middle can look similar even when their values differ.

An authentic recording can still be edited into a false sequence
Real audio or video can be rearranged, shortened or separated from context until it appears to tell a story the full recording does not support.

A rolling average smooths noise—and delays turning points
A moving window makes a noisy series easier to read, but the smoother line responds later when the underlying direction changes.

A password manager is only as secure as its recovery plan
A strong vault needs a safe route back in after a lost device or forgotten passphrase—without creating an easier route for an intruder.

A meaningful right of reply needs time and specific questions
Seeking comment is more than sending a last-minute message. The subject must understand the important claims and have a realistic chance to answer.

A logarithmic chart compresses equal percentage changes
On a logarithmic axis, equal distances represent equal ratios rather than equal amounts—a useful view that changes how growth appears.

An official statement is a source, not independent proof
Statements show what an institution says happened. Reporting must still test the facts, context and competing accounts.

Selection bias: who is missing from the result?
A large dataset can still mislead when the people included differ from the people the conclusion describes.

Public Wi-Fi and HTTPS protect different things
Encrypted websites reduce what nearby observers can read, but the network and the device still require careful choices.

Naming a victim requires a public-interest test
The fact that a name is available does not always mean publishing it will improve the public’s understanding.

Cumulative totals and daily changes tell different stories
One shows everything recorded so far. The other shows how quickly the total is moving now.

A document screenshot needs the original file
A cropped page can hide headers, footnotes and surrounding language that determine what the document means.

Median and percentile answer different questions
Both describe position in a distribution, but one marks the middle while the other shows how a value compares with the rest.

Review app permissions before they become permanent
Temporary needs can leave lasting access to location, contacts, photos and microphones long after a task ends.

Conflict disclosures should be close to the claim
Readers need relevant relationships before they evaluate an argument, not after they have already accepted it.

A rate needs a time period
Per person is only part of the denominator. The length of observation can change the comparison completely.

How to tell a press release from independent reporting
A polished announcement can resemble a news story, but the source, purpose and unanswered questions are different.

A zero in a dataset may mean missing
Nothing happened and nothing was recorded are not the same result. Data codes can hide that difference.

Two-factor authentication can still be phished
A second step improves account security, but fake sign-in pages and approval fatigue can still defeat weaker methods.

Why editor’s notes belong on sensitive stories
A short note can explain an unusual decision, a reporting limit or a risk that ordinary article labels cannot show.

Rankings hide the choices that built them
A single position can combine dozens of measurements, weights and missing values into one deceptively simple number.

Find the primary document behind the headline
A report about a report is useful, but the original document shows what was actually measured, decided or said.

Relative risk can make a small change sound huge
A percentage increase needs its starting point. Absolute numbers show how many people or cases are actually affected.

Password reuse turns one breach into many
When the same password protects several accounts, a leak from one service can unlock doors elsewhere.

A byline is not the same as a dateline
One identifies who reported the story. The other traditionally tells readers where that reporting was filed.

Preliminary data is designed to change
Early estimates provide speed. Revisions provide completeness. The label tells readers which tradeoff they are seeing.

How to read an eyewitness video without overclaiming
A video may show something real while leaving the location, timing and wider event unresolved.

Base rates: the number that changes the odds
A result can look highly accurate and still produce surprising mistakes when the underlying event is rare.

Private browsing does not make you invisible
Private mode limits what stays on your device, but it does not erase every record created beyond it.

What anonymous sourcing should tell the reader
Protecting a source may be necessary, but the article still owes readers enough context to judge the claim.

How survey wording can change an answer
Small changes in order, tone and response options can move a survey result before anyone answers.

How to check a viral quote before you share it
A quote card can travel farther than its original context. Four checks can show whether the words, speaker and meaning still line up.

Why small samples create fragile headlines
A dramatic percentage can rest on very few observations. Sample size helps reveal how much confidence a result deserves.

What photo metadata can—and cannot—prove
File details can support an investigation, but missing or editable metadata should never carry the whole argument.

A news update is not the same as a correction
New facts and repaired errors serve different purposes. Clear labels help readers understand how a story changed.

How a map can change the story
Scale, projection, boundaries and labels shape what a map appears to say. Read the design before accepting the message.
Watch & understand
All videos →A clearer picture, one explanation at a time. Selected videos from TED-Ed.

How to spot a misleading graph
TED-Ed · Lea Gaslowitz
Watch on YouTube ↗
How to choose your news
TED-Ed · Damon Brown
Watch on YouTube ↗
How false news can spread
TED-Ed · Noah Tavlin
Watch on YouTube ↗News literacy
View all →The numbers
View all →Technology
View all →Editorial perspective
View all →Clarity in a fast-moving world.
101 original articles. Practical explanations. Transparent perspectives.
Read our editorial standards →





