How to Spot Fake News
A practical, checkable guide you can run on any article in a few minutes — no special tools required, though we’ll point to one for each step that has one.
“Fake news” covers a wide range of problems: stories that are fabricated outright, real events reported with a framing the facts don’t support, quotes used out of context, and claims that change between a headline and the article underneath it without the change being noted. Most of it isn’t something you have to take on faith. This guide walks through six checks a reader can do on their own, and links to the Normalize the News tool that runs the matching check and shows its evidence. Here’s how to check it yourself, with receipts.
1. Check the primary source
Before evaluating how a story is written, find out where its central claim comes from. Look for the specific document, statement, dataset, or original report the article is describing — a press release, a court filing, a government report, a study, or another outlet’s first-hand reporting. Phrases like “officials said” or “a new report found” are pointers, not evidence on their own. If the article links to the source, open it. If it doesn’t, search for it directly — most primary sources (agency statements, court dockets, academic papers) are public and searchable. If you can’t find a primary source that says what the article says it says, that’s useful information by itself.
2. Compare the headline to the body
Headlines are written to get attention, and they’re sometimes written by a different person than the one who wrote the article. That gap is where a lot of distortion happens: a headline can state something more strongly, more broadly, or more definitively than the reporting underneath it actually supports. Read the headline as its own claim, then read the body specifically looking for the sentence that supports or contradicts it. If you can’t find that sentence, the headline is making a claim the article doesn’t back up.
NTN’s Headline vs. Body tool does this automatically: for each tracked article, it extracts the claim the headline is making and shows the exact sentence from the body that does or doesn’t support it, along with a confidence level. It doesn’t label a headline “misleading” — it shows you both pieces of text side by side so you can compare them yourself.
3. Look for the full quote and its context
A shortened quote can say something different from what the full quote says, even without changing a single word inside the quotation marks. When an article quotes someone, check whether the sentence before and after the quoted portion is included. If multiple outlets covered the same statement, compare how much of it each one printed. A quote that ends right before a qualifying clause, or that starts right after an important piece of context, is worth a second look.
NTN’s Quote-omission tool flags cases where a quote reported in full by one outlet appears in a shortened form in another outlet’s account of the same story, so you can compare the full version against the trimmed one directly.
4. See who else is — and isn’t — covering it
A single outlet reporting a claim is one data point. Checking whether other outlets are covering the same story, and how they describe it, gives you more to work with. Coverage that’s concentrated in a small number of outlets, or that’s reported very differently across outlets that normally cover similar ground, is worth noting — not as proof of anything on its own, but as a signal to look closer at the primary source.
NTN’s Coverage ledger keeps a record of which outlets in its tracked set did, and did not, cover a given story, so you can see the spread of coverage in one place instead of searching outlet by outlet.
5. Check whether the article changed after it was published
Online articles get updated after they go live — sometimes for a typo, sometimes to add new information, and sometimes in ways that change a fact, a figure, or a quote. Most outlets don’t flag every change on the page itself, so the only way to know is to compare the version you’re reading against an earlier capture of the same URL. If a specific number, name, or claim seems important, it’s worth checking whether it read differently at an earlier point.
This is what NTN’s Revision Tracker is built for: it tracks articles and automatically publishes a change once it appears in at least two confirmed Internet Archive captures and clears a fixed noise threshold — no human decides what to show. Every published revision shows exactly what the text said before and what it says now, and links both public captures so you can verify it yourself.
6. Check provenance and authorship
Who wrote the piece, and is there any record of where a published photo or video actually came from? A byline, an “About the author” page, or a stated affiliation gives you somewhere to start. For images, a growing number of publishers attach a Content Credential (a C2PA-standard record of an image’s edit history) — when it’s present, you can check it; when it’s absent, that tells you the signal isn’t available, not that anything is wrong with the image.
NTN’s Provenance tool checks published images for a Content Credential, checks for AI-disclosure or sponsored-content labeling on the page, and notes whether the outlet appears in a journalism trust registry — all surfaced as neutral, checkable signals rather than a score.
Putting it together
No single one of these six checks proves a story is accurate or inaccurate on its own. Together, they replace a gut reaction with something you can actually point to: a primary source you looked up yourself, a headline you compared against the body text, a quote you read in full, a sense of how widely a claim was covered, a record of whether the article changed, and whatever provenance information is available. That’s the same approach behind every NTN tool — publish the evidence and a confidence level, and let the reader draw the conclusion. For the exact rules each tool follows, see the methodology page.