Start with the source
Before weighing what a story says, establish who is saying it. Most bad information fails this first test, and most people skip it.
The distance rule
Information degrades as it travels. An agency report is more reliable than a story about the report, which is more reliable than a post summarizing the story, which is more reliable than a screenshot of that post. Every hop adds compression, framing, and error. When a claim matters to you, walk it back toward its origin.
- Primary source The record itself: official statements, filings, data, direct footage, a firsthand account. Strongest evidence, though not automatically true.
- Secondary source Reporting that gathers and interprets primary sources. Quality varies with the outlet's sourcing discipline.
- Tertiary echo Aggregators, reaction posts, screenshots, and summaries of summaries. Never accept these as evidence on their own.
Questions that expose weak sources
- Does this outlet publish corrections, or do errors just disappear?
- Does the story name its sources, or does it lean on "reports say"?
- Does the author have a track record on this beat?
- Is there a working way to contact the publication and a stated editorial policy?
- Who funds the outlet, and would this story look different if the funder changed?
Framing and bias
Bias is rarely fabrication. Far more often it is selection: which facts get included, which get left out, and which get the headline. A story can be entirely accurate and still leave you with a false impression.
Four moves to watch for
- Framing. The same event can be "protesters clash with police" or "police clash with protesters." Notice who gets the verb.
- Omission. What context would change how this reads? A statistic without its baseline, a quote without its question, an event without its history.
- Emphasis. Was the concession buried in paragraph nine? Placement is an editorial decision, and it shapes what most readers take away.
- Loaded language. "Admitted" versus "said." "Regime" versus "government." Word choices carry verdicts.
Your own bias is part of the system
Confirmation bias means stories that flatter your existing views get less scrutiny from you, not more. The practical fix is mechanical, not moral: apply the most skepticism to the stories you most want to be true, because those are the ones you will share before checking.
Reading images and video
Most misleading imagery is not fabricated - it is real footage wearing the wrong caption. An earthquake clip from 2011 resurfaces for every new earthquake. The second-largest category is selective footage: real, current, but cropped in time or space to imply something that did not happen.
- Ask when before what. A reverse image search often reveals a "breaking" photo is years old.
- Check the setting. Signage, weather, vegetation, license plates, and architecture should match the claimed place and season.
- Watch the edges. What is just outside the crop? When did the clip start rolling? The moments before a dramatic clip often change its meaning.
- Treat AI-era media with baseline suspicion. Synthetic images and video are now cheap. Hands, text, reflections, and physics still fail often - but the reliable test is provenance, not pixel-hunting.
The more emotionally powerful an image is, the more valuable it is to people who want to mislead you - and the more it deserves a provenance check before you believe or share it.
Numbers and statistics
Numbers feel objective, which is exactly why they are useful for misleading people. A few questions catch most statistical sleight of hand:
- Compared to what? "Cases doubled" means nothing without the base. Two cases becoming four is a doubling.
- Percent of what? A "50% increase in risk" from 2-in-a-million to 3-in-a-million is real and also nearly irrelevant.
- Who counted, and how? Definitions drive numbers. "Unemployment" or "casualties" can each be counted several defensible ways with very different results.
- Is the axis honest? Truncated axes and cherry-picked date ranges make flat lines dramatic and trends invisible.
- Correlation or cause? Two lines moving together is the beginning of a question, not the end of one.
How misinformation works
Misinformation is not random noise; it follows patterns because the patterns work. Knowing them is most of the defense.
- Emotional overload. Content designed to trigger fury or fear first, because aroused readers share before they check.
- False urgency. "Before this gets deleted" and "the media won't show you this" are pressure tactics, not sourcing.
- Borrowed credibility. Fake quotes attached to real names, real logos on fabricated headlines, lookalike domains one letter off.
- The kernel of truth. The most effective false stories start from a real event and bend it - which lets them survive a lazy fact-check.
- Unfalsifiable framing. If every possible piece of evidence somehow confirms the theory, including the debunkings, you are looking at a belief system, not a claim.
- Just asking questions. Insinuation does the work while the author keeps deniability. Notice when a story implies what it never states.
Habits that hold up
Media literacy fails as a one-time lesson and works as a routine. These five habits cover most situations:
-
01
Pause before sharing
The share button is where misinformation wins or loses. If a story spikes your pulse, that is the moment to slow down - you are the target audience for a reason.
-
02
Read laterally
Professional fact-checkers don't inspect a suspicious site; they leave it. Open new tabs, search the claim, and see what independent sources say about it and about the outlet.
-
03
Trace to the original
For any claim that matters, find the primary source: the actual study, the full statement, the unedited video. Summaries mutate.
-
04
Separate developing from confirmed
Early reports of any breaking event are partially wrong, everywhere, always. Hold early details loosely and expect the record to move. This is why our own updates are labeled the way they are.
-
05
Diversify deliberately
Read outlets that disagree with your defaults - not to split the difference, but so selection bias in any one feed stops being invisible to you.
Tools worth knowing
A small toolkit covers most everyday verification:
- Reverse image search. Google Lens and TinEye reveal where an image appeared before and when.
- Fact-check archives. Snopes, Reuters Fact Check, AP Fact Check, and FactCheck.org have often already traced the exact claim in your feed.
- Page archives. The Wayback Machine and archive.today show what a page said before it was edited or deleted.
- Primary data. USGS for earthquakes, the National Weather Service for storms, court records for legal claims - the same feeds our newsroom reads directly.
For the step-by-step checks we run on every story - and that you can run yourself - see Fact-checking tips. To practice against real examples, the Fact-Checker Game uses actual headlines.
Put it into practice.
Real headlines, real stakes. Call verified or false, then see the evidence and how the check was done.
News on social platforms
Social feeds are optimized for engagement, and outrage engages. That single fact explains most of what goes wrong with news on platforms: the version of a story that spreads fastest is the version engineered to make you feel, not to inform you.
The strongest single habit: before resharing anything consequential, open it. Read past the headline. Check the date. A large share of viral news is old news wearing a new outrage.