How to Verify Something You See Online

STOP

Before reacting, sharing, paying, clicking or panicking:

Stop.

This matters especially when something creates a strong emotional reaction.

Maybe you're:

• angry
• scared
• excited
• shocked
• disgusted
• desperate to help
• desperate to warn someone

You don't have to decide immediately whether something is true.

Sometimes the smartest first move is simply:

Don't share it yet.

SOURCE

Ask:

Where did this actually come from?

Not:

• Who sent it to me?
• Who reposted it?
• Which Facebook group did I see it in?

Try to find the original source.

If it's a screenshot, can you find the original post?

If it's a quotation, can you find the original speech, interview or document?

If it's a video, can you find the full recording?

If it's a statistic, who collected the data?

If it's an AI answer, what source supports the claim?

SEARCH

Now look somewhere else.

This is sometimes called lateral reading.

Instead of spending all your time examining one website, post or video, open another search and investigate it from outside.

The Digital Inquiry Group's Civic Online Reasoning materials specifically teach lateral reading and other research-backed strategies for evaluating online information.

Search:

• the claim
• the person's name
• a distinctive sentence from the post
• the supposed event
• the organization
• the image
• the statistic
• the quotation

You may discover that the information is accurate.

You may discover that it's false.

Or you may discover something very common:

It's missing context.

CONFIRM

Now ask:

What evidence actually supports this?

Look for confirmation outside the original claim.

Depending on the situation, that might mean:

• an official agency
• the organization supposedly involved
• an original document
• multiple credible news organizations
• the full video
• an original research paper
• a professional source
• another independent witness or recording

One source repeating another source isn't necessarily independent confirmation.

Try to trace information back toward where it originated.

Not automatically. Multiple search results may all be reporting the same original source.

Example:

• Website A reports an original story
• Website B copies the story from Website A
• Website C copies the story from Website B
• Your search results show all three

That's still ONE underlying source repeated multiple times, not three independent sources.

When you look at search results, watch for phrases like:

• "according to..."
• "reported by..."
• "citing..."
• "based on..."
• "as stated in..."

These phrases often trace information backward to its origin.

Key principle: Repetition is not independent confirmation.

The goal is finding evidence that actually originates from different places, not just finding more places that say the same thing.

The right source depends on what you're checking.

Health or medical information

• Established health agencies (WHO, CDC, national health ministries)
• Medical organizations in relevant specialties
• Qualified medical sources
• Original research when appropriate

Government rule, program, or service

• The responsible government agency
• Official documentation
• Current information from official sources

Company, product, or service feature

• Current official company documentation
• What the company actually says today
• Official product pages or announcements

School, university, or local institution

• The institution itself
• Responsible school districts or municipalities
• Relevant government agencies

News or current event

• Credible reporting, preferably independent
• Multiple credible news sources
• Reporting that traces back to evidence or sources

Scientific or research claim

• The original study or research
• Qualified expert interpretation
• Research institutions

AI answer or summary

• Evidence OUTSIDE the AI answer itself
• Sources that support the claim
• Real-world verification, not the AI's confidence

The principle is: Who or what would reasonably be in a position to know this?

Reliable information can become outdated.

Ask:

• When was this published?
• When did the event happen?
• Has the policy, product feature, or recommendation changed since then?
• Is newer information available?

Example: A reliable article about a government program from five years ago might accurately describe what was true then but be wrong about what is true now.

For claims where updates matter—policies, products, medical guidelines, technology features—look for current information from the responsible source.

Understanding what kind of information you're looking at helps you make better decisions.

Evidence: Something that directly supports or shows what was said happened. Data, recordings, witnesses, documents, research results.

Opinion: Someone's interpretation, judgment, or belief. Usually introduced with phrases like "I think," "In my view," "This shows that..." when the connection is the person's judgment rather than proven fact.

Prediction: What someone thinks might happen. "Experts predict," "If this continues," "This could lead to..."

Repetition: The same claim being said again without new evidence. Useful for confirming a claim has been widely reported, but not the same as independent sources reporting evidence.

None of these types of information are automatically useless. The point is knowing what kind you're looking at.

For an important decision, look for evidence. For background, opinions can be useful. For preparation, predictions can matter. But don't confuse any of these with proof.

Disagreement among credible sources doesn't mean "nobody knows anything." It may mean:

• Evidence is incomplete or still developing
• The evidence is being interpreted differently
• They are actually answering slightly different questions
• New information has emerged that changes the answer

When credible sources disagree:

1. Check whether they are actually answering the same question
2. Check dates—has something changed?
3. Look at what evidence each source uses
4. Look for areas where they DO agree
5. For important decisions, seek appropriate qualified expertise or wait for clearer information when possible

For your own decisions: It's reasonable to say "The evidence here is mixed" or "I'll wait for more information" rather than forcing certainty where experts themselves disagree.

You are not required to prove something false before deciding not to trust, act on, or share it.

Reasonable choices when you can't verify something:

• Don't share it yet. It's okay to see something and decide not to pass it on
• Don't make a high-stakes decision from it. Wait for clearer information if the decision matters
• Say "I couldn't verify this." You can tell others you looked but couldn't confirm it
• Look again later if information may change. Some claims become clearer over time
• Seek an appropriate source if it matters. For health, legal, financial, or important personal decisions, reach out to a qualified source instead

"I don't know yet" is a perfectly valid conclusion.

It's better than confidently sharing or acting on unverified information.

A screenshot is not a source.

Screenshots can be:

• real
• edited
• cropped
• generated
• recreated
• old
• missing context
• attached to a false explanation

If a screenshot matters:

Try to find the thing being screenshotted.

A real photo can support a false story.

The photo may be authentic.

The caption may not be.

Old photos and videos can be reposted as if they happened somewhere else or at another time.

AI-generated doesn't automatically mean false.

Someone can use AI to create an illustration that is clearly labeled and completely honest about what it is.

Human-created doesn't automatically mean true.

People have created misinformation, propaganda, scams, fake quotations and manipulated media for much longer than generative AI has existed.

The creator being human doesn't verify the claim.

They can help.

But they shouldn't be your entire verification process.

NIST's current deepfake work shows why detection performance can decline substantially under real-world conditions.

Instead of asking only:

"Was this made by AI?"

also ask:

"Is the claim true?"

Those are not the same question.

Content Credentials may provide useful information about the origin or editing history of some digital media.

They offer a kind of digital history that can show where media came from and what changes were made to it.

C2PA specifically explains that provenance information alone cannot tell you whether an image or video depicts something true, accurate or factual. It also warns that the absence of Content Credentials should not mean content should automatically be distrusted.

Think of provenance as evidence, not a truth label.

You do not need to investigate every meme, recipe or silly video you encounter.

Think about the consequence of being wrong.

Before you:

• send money
• accuse someone
• share something alarming
• make a health decision
• make a financial decision
• repeat a damaging claim
• act on an emergency
• make an important decision

...verification matters more.

And sometimes you won't be able to determine whether something is true.

That's okay.

"I don't know yet" is a perfectly reasonable conclusion.

You don't have to prove something is fake before deciding not to trust, act on or share it.

Remember: