How to Fact-Check a News Story: Methods and Fact-Checking Tools

To fact-check a news story, you compare its claims with independent sources before you believe or share it, and six checks done in the same order are usually enough. A few free fact-checking tools make each check faster. A false story shared by a company account costs more than ten missed posts. You don't need a verification desk to avoid that, just a two-minute routine you run before you hit publish.

Below you'll find the routine step by step and the free tools that speed it up. At the end there's a section on where AI actually helps, which is often not where people expect.

What fact-checking is

Fact-checking means verifying a factual claim against independent sources. It covers facts and leaves opinions out. "The government allocated 200 million" can be checked. "The government made the wrong call" can't.

It's related to two other jobs. Debunking takes apart a false story that is already circulating, and source verification looks at how reliable the publisher is rather than at a single fact. Most people do all three in daily work without noticing.

Six checks, in the right order

Order matters. The first two checks rule out most cases in a few seconds, and only what survives them deserves the other four.

1. Who is saying it

Go back to the original source and ignore whoever is passing it along. If a post cites "a study", find the study. If it cites "sources close to the matter", you're reading a leak that still needs confirming.

The most common warning sign is a chain. One site picks up another site, which picked up a post, and at the start there's nothing. If after three hops you can't find a document, a news agency or a named person, the story doesn't exist yet.

2. When

The publication date catches more hoaxes than any other check. A lot of false news is real news from years ago, pushed back into circulation as if it happened today. The flood or the arrest really did happen, only it was in 2019.

Check the date on the article. If there isn't one, look at the URL or search the headline in quotes to see when it first appeared. The Wayback Machine keeps archived copies of many pages, so you can see what an article looked like on a given date.

3. What others are saying

This is the most effective technique, and it's called lateral reading. Instead of studying the page in front of you, you leave it and look up what others say about the same fact. In a Stanford History Education Group study, professional fact-checkers left a site after a quick scan and opened new tabs to check it. Compared with historians and students, they reached better-founded conclusions in much less time.

In practice, take the central claim, search for it on Google and see whether news agencies or independent outlets report it. If a significant fact comes from a single source, ask yourself why nobody else is covering it.

When only one outlet runs an important story, chances are the others have already checked it and passed.

4. Images

Photos are the weak spot of most false stories, because reusing an old image is much easier than building a convincing new one.

Run a reverse image search with Google's search by image or with TinEye. If the same photo shows up in articles from three years ago, you're done. For videos, grab a frame and search for it the same way. The InVID-WeVerify plugin, built for journalists, splits videos into keyframes, runs reverse image searches and reads metadata.

With AI-generated images the old giveaways, like mangled hands, work less and less. Other clues still help, such as inconsistent shadows, unreadable text on signs or background details that repeat. And the earlier question still applies: has anyone else documented the same event? We cover file provenance, watermarks and deepfake detectors in our guide to verifying media integrity with AI.

5. Numbers

Figures get checked at the source, which is almost always public (national statistics offices, Eurostat, company filings, council resolutions). The most common trick is stripping a real number of its context. "Cases tripled" can mean going from two to six.

Always check the absolute value next to the percentage, the time period and who paid for the research.

6. What fact-checkers have found

If doubt remains, see whether someone has already done the work. Internationally, Reuters Fact Check, AFP Fact Check and Snopes publish checks regularly. In Italy the steady names are Facta, Pagella Politica, BUTAC and Paolo Attivissimo's Il Disinformatico.

Google Fact Check Explorer searches fact-checks that have already been published, by keyword or by image, and shows the rating and who gave it. Thirty seconds there often settles the question.

flowchart TD
    A["Story to check"] --> B{"Can you find the original source?"}
    B -->|"No"| X["Don't publish"]
    B -->|"Yes"| C{"Is the date recent?"}
    C -->|"No"| X
    C -->|"Yes"| D{"Do other independent sources report it?"}
    D -->|"No"| E["Flag as unverified"]
    D -->|"Yes"| F{"Do the images and numbers hold up?"}
    F -->|"No"| X
    F -->|"Yes"| G["OK to publish, with attribution"]

Free fact-checking tools

A handful of free fact-checking tools covers all six checks. The table lists the main ones and the check each one helps with.

None of these tools gives you the answer on its own. They speed up a check that is still yours to make.

Where AI helps, and where it doesn't

The most common mistake is asking a chatbot "is this story true?". A general-purpose language model often doesn't consult up-to-date sources, and it tends to confuse authority with how often something gets cited. In the worst cases it invents a confirmation that doesn't exist. It's the wrong tool for this job.

AI works better upstream, by cutting down the number of doubtful stories that land on your desk. A system that automatically compares a claim with several independent sources does the third check on this list for you. That's lateral reading, the check that takes the longest by hand. The methods behind these systems are covered in our guide to AI fact-checking techniques.

Scova AI has a fact-check feature that compares a story with authoritative sources and returns one of four verdicts.

The last row matters most and gets misread most. An unverifiable story can turn out true or false, and for now there aren't enough sources to tell. Usually it's too fresh, too niche or has no secondary coverage. If you publish on social media, that means waiting a few hours. If you're a journalist, it means picking up the phone before you use it.

Making it part of daily work

Verification only lasts if it doesn't feel like an extra step. Three habits keep it sustainable.

Run the checks while you read, before you write the post. Checking finished content costs twice as much, because by then you've grown attached to it.

Cut the noise upstream. A news feed built on sources you picked contains far fewer things to verify than a general one. That's half the work of good media monitoring, and it's why Google Alerts alone isn't enough.

Write the rule down. An agency handling several clients needs one written line saying which content can go out without checks, which needs a second person and who decides in unclear cases. Without that line, the call falls on whoever is on shift.

FAQ

What is fact-checking?

It's verifying a factual claim against independent sources before you believe or share it. It applies to facts such as numbers, dates and statements, and leaves opinions out.

What are the best free fact-checking tools?

Google Fact Check Explorer for checks that are already published, Google search by image and TinEye for photos, the InVID-WeVerify plugin for videos and the Wayback Machine for dates and old versions of a page. They cover the six checks in this guide.

How long does it take to fact-check a story?

Two minutes in most cases. The first two checks, source and date, catch a large share of false stories. It only stretches to ten or fifteen minutes when the content is sensitive and it's worth calling someone.

How can you spot a fake news story?

The most frequent signs are an original source you can't find, an old date, no other outlet reporting it, images that already appeared elsewhere with a different story, and numbers without context. If you find even one, verify before you share.

Which fact-checking websites are the most reliable?

Internationally, Reuters Fact Check, AFP Fact Check and Snopes are steady references. In Italy the most established are Facta, Pagella Politica, BUTAC and Paolo Attivissimo's Il Disinformatico. Google Fact Check Explorer lets you search checks from many fact-checkers in one go.

How do I recognize an AI-generated image?

Visual clues are less and less reliable. Two indirect routes work better. A reverse image search often turns up the original or the context, and a real event is usually documented from several angles and by several sources.

Are AI content detectors reliable?

Not very, so treat them as a clue. False-positive rates are high, especially on translated text or text written by non-native speakers. Content written with AI isn't necessarily false, and content written by a person isn't necessarily true.

What should I do if I've already published a false story?

Correct it quickly and visibly. Edit or delete the content, post a correction on the same channel with the same prominence and explain what happened. A fast correction does far less damage than a quiet deletion that someone has already noticed.

The takeaway

Fact-checking makes publishing repeatable. People who work without a method swing between weeks when they check everything and weeks when they check nothing, and sooner or later they pay for it. Six checks done the same way every time cost less than a correction, and a lot less than rebuilding a reputation.