The SIFT method

Why it matters

Most of what you see in a feed is not worth a deep investigation. It is also not worth believing on sight. The problem is the gap between those two: you do not have twenty minutes per post, so you end up judging by gut (does it look legit, does it feel right?), which is exactly the sense that manipulation is aimed at.

SIFT closes that gap. Mike Caulfield, who teaches digital literacy, built it in 2019 out of what professional fact-checkers actually do: not reading more closely, but leaving the page and checking around it. Four moves, in order, usually under a minute. The point is not to prove something false. It is to find out, fast, whether this thing deserves any more of your attention at all.

How to do it

1. Stop. Before you react, reply or share, notice what you are feeling. Anger, fear, glee, “I knew it”: strong emotion is a signal to slow down, not a reason to hit share. Then ask two things: do I know this source, and do I know this claim? If either answer is no, keep going.
2. Investigate the source. Not by reading the post more carefully, but by leaving it. Search the name of the site or account and add “wikipedia” or “who owns”. You want what other people say about the source, not what it says about itself (see lateral reading). Thirty seconds is usually enough to tell a newsroom from a three-month-old site with no address.
3. Find better coverage. Take the claim itself (not the link) and search for it. If a story is big, someone with a reputation to lose will have covered it. If the only places reporting “the leaked document” are five sites copying each other, that tells you something. If a fact-checker has already handled it, you are done.
4. Trace to the original. Quotes, numbers, photos and “studies” almost always arrive second- or third-hand. Click through until you reach the original document, speech, video or paper, then read what it actually says. Very often the original is real and the summary is the fiction (see misleading headlines and out-of-context media).

You can stop at any step. If step 2 shows a reliable source and step 3 shows others reporting the same thing, you do not need step 4. If step 2 shows the source is nobody, you may not need anything else.

Example

@eurofreedom_news, 4h LEAKED: Brussels document reveals plan to BAN wood-burning stoves in all homes by 2027. They want you cold and dependent. Read before it’s removed: euro-freedom-report.com/stoves

Stop. “They want you cold”, “read before it’s removed”: the post works on your fear and your sense of urgency before it tells you anything. That is the cue.

Investigate. A search for “euro-freedom-report” finds no newsroom, no editor, no address, and a domain registered a few months ago. Not proof of anything, but it puts the burden of proof squarely back on the post.

Find better coverage. Searching “EU wood stove ban” turns up explanations from national newspapers and fact-checkers: the EU’s ecodesign rules set emission limits for new stoves sold since 2022. Nobody is removing stoves from homes.

Trace. The “leaked document” link leads to a public consultation PDF that has sat on the Commission’s website since 2022. Public is the opposite of leaked.

About ninety seconds, and the post has gone from “alarming” to “a real rule, wrapped in a fiction”. Note what you did not have to do: argue, know anything about stoves, or decide who to trust in general.

Common mistakes

• Reading harder instead of leaving. Staring at the post for signs of fakery is what the post was designed for. The information you need is elsewhere.
• Trusting the “About” page. It was written by the people you are checking. Would a scam site describe itself as one?
• Stopping at the first match. Five sites repeating the same wording is one source, copied, not five confirmations. Look for someone who actually did the reporting.
• Treating SIFT as a verdict. It sorts; it does not prove. “I couldn’t find the original” is a reason not to share, not a reason to call someone a liar in the comments.
• Skipping the Stop. The other three moves do not happen if you have already shared. Most of the method’s value is in the first three seconds (see pause before sharing).