A Four-Step Protocol for Verifying a Source

⚠️ Version note: the interfaces named here (reverse image search, fact-check sites) move around; the four moves don't. If a menu isn't where this says, check the "If it moved" note under that step. Last checked: 25 July 2026.

The people who get fooled are the ones who read carefully

The instinct, when you hit a claim you're unsure about, is to read it more closely — study the page, weigh how professional it looks, read to the bottom. This is exactly the wrong move, and research on how professional fact-checkers work shows why: the experts spend almost no time on the page itself.

Fact-checkers read laterally — they leave the page immediately and open new tabs to find out what other sources say about it. Students (and most adults) read vertically — they stay on the page and try to judge it from the inside, which is precisely how a well-designed misleading page wins: it controls everything you can see without leaving.

This lesson is a four-step protocol you can run in under two minutes on anything — a headline, a screenshot, a viral clip, a statistic, an AI answer. It's adapted from Mike Caulfield's SIFT, and the output is your own checklist you'll actually use. The point isn't to verify everything. It's to spend thirty seconds before you believe or share.

Before anything else: notice your reaction and pause.

The single strongest signal that something needs checking is that it made you feel something instantly — outrage, vindication, fear, delight. Content engineered to spread is engineered to trip exactly that reaction, because emotion is what makes you share before you think. The stronger the pull to react, the more it's worth ten seconds.

Two concrete stops:

• Do not share yet. Sharing is endorsing, and you can't un-ring it. Verify first, share second.
• Check your own bias. If a claim flatters what you already believe, you are the easiest possible mark for it — you'll want it to be true. Notice that want.

✅ Check: you can name the emotion the content triggered, and you haven't shared or repeated it yet.

Not is this source good? — that's the vertical trap. Instead: who is behind this, according to everyone except them?

• Open a new tab and search the source's name. Read what independent sources say about the outlet, the account, the organisation. A site's own "About" page is a press release; other people's descriptions are information.
• Ask what kind of thing it is. A peer-reviewed study, a newspaper report, a company's own press release, an anonymous account, a satire site, an ad dressed as an article? The type tells you what it's built to do before you read a word of the content.
• Check the expertise and the stake. Is the person qualified on this topic, and do they gain something (money, attention, a political win) if you believe it?

You are not deciding "trustworthy / untrustworthy". You're placing the source: what is this, who runs it, what do they want.

If it moved: you don't need a special tool for this — a plain web search of the outlet or account name, plus the word "owner", "funding", or "bias", does most of the work.

✅ Check: in one sentence, you can say what the source is and what it stands to gain.

For most claims you don't need to evaluate this source at all — you need to find out what the best sources say about the claim, and read those instead.

• Search the claim itself, not the page. Type the core assertion into a search engine and see who else is reporting it and how.
• Look for coverage that disagrees or adds nuance. If only the original source and its copies carry the claim, and no independent outlet does, that's a finding. If reputable sources report it with important caveats the original dropped, that's the real story.
• A single high-quality source beats ten low-quality ones. Ten sites repeating each other is one source wearing ten coats.

The move is lateral: trade this page for a better page on the same claim.

✅ Check: you've found at least one independent source on the claim — or confirmed that suspiciously nobody else is carrying it.

Claims, quotes, images and clips get worse every time they're re-shared — stripped of context, cropped, re-captioned, misdated. The fix is to go upstream to the source of the source.

• Quotes and statistics: find where they originally came from. A "study says" with no link to the study is a red flag; a stat with no origin is a rumour with a number on it.
• Images and video: run a reverse image search (upload the image, or a screenshot of one frame, to a reverse-image tool) to find where it first appeared. Half of "shocking" viral photos are real photos from a different event, place, or year — the image is genuine, the caption is the lie.
• Read past the headline. Often the article the headline sits on top of directly contradicts or heavily qualifies it. Headlines are written to be clicked, not to be accurate.

If it moved: most browsers and search engines have a "search by image" option (a small camera icon in the search bar, or right-click an image → "search image"). If yours doesn't, any dedicated reverse-image site works the same way.

✅ Check: for the key claim, quote or image, you've reached the original context — or confirmed you can't, which is itself a reason to distrust it.

Assemble your checklist

Everything above is four moves. Write them where you'll see them:

BEFORE I BELIEVE OR SHARE:
[ ] STOP — what am I feeling, and does it flatter what I already think?
[ ] SOURCE — what is this, and what does it gain if I believe it?
[ ] BETTER COVERAGE — what do independent sources say about the claim?
[ ] ORIGINAL — traced the quote / stat / image to where it started?

That's the reusable core. Most checks stop at step 2 or 3 — you rarely need all four. The discipline is running any of them before you react.

What this means for you

• Read laterally, not vertically. Leave the page and see what others say about it. Judging a page from the inside is how good fakes win.
• The strongest verify-me signal is your own reaction. Outrage and delight are the tells.
• A source's "About" page is marketing. What others say about it is evidence.
• Ten copies is one source. Volume isn't corroboration; independence is.
• Real image, false caption is the most common visual lie. Reverse-search before you believe.

Try it: verify three real claims (30 min)

1. Take three things you actually saw this week — a viral post, a stat someone quoted, a screenshot, a headline, an AI-generated answer you're unsure about.
2. Run the four moves on each and write:

CLAIM / ITEM: .................................................
STOP: what it made me feel .....................................
SOURCE: what it is + its stake .................................
BETTER COVERAGE: what independent sources say ..................
ORIGINAL: where the quote/stat/image really came from ..........
VERDICT: solid / misleading / can't tell — and why ............

3. For at least one image or clip, do a real reverse image search and note what you found.

✅ Finish check: three items run through all four moves, each with a verdict and a reason, and at least one reverse image search actually performed.

Summary card

• Read laterally: leave the page, open new tabs, ask what others say. Don't judge from the inside.
• STOP — notice the emotion; don't share yet; watch your own bias.
• SOURCE — what is it, and what does it gain?
• BETTER COVERAGE — search the claim; find independent sources; ten copies = one source.
• ORIGINAL — trace quotes, stats and images upstream; reverse-search images; read past the headline.
• You rarely need all four. The discipline is running one before you react.

Sources

1. Caulfield, M. — SIFT (The Four Moves), 2019
2. Wineburg, S. & McGrew, S. — Lateral Reading, Stanford History Education Group, 2017
3. Wardle, C. — Information Disorder, First Draft, 2017

Next lesson: DT-07 — Ten Cognitive Biases and Where You'll Meet Them Today (L1) Related: DT-03 Correlation Is Not Causation · AI-07 Hallucination Hunting · MN-10 The Anatomy of a Scam · DT-02 How Charts Lie Path: Hard to Fool — 4/6