How do skilled people decide whether to trust something online? Researchers at the Stanford History Education Group set out to find out. In a study first released in 2017, they asked three groups to evaluate websites and claims: professional fact-checkers, historians with doctorates and Stanford undergraduates. The fact-checkers were far faster and more accurate than both other groups.

Lateral reading

The difference was in how they read. The historians and students mostly read vertically: they stayed on the page, studied its design, read its “About” section and judged it by how professional it looked. But a polished website and an official-sounding name are easy to fake.

The fact-checkers read laterally. Within seconds of landing on an unfamiliar site, they left it, opened new browser tabs and searched for what other sources said about the site and its claims. They learned more about a source by reading about it than by reading it. They also practiced click restraint: scanning search results before choosing which to open, instead of clicking the first link.

SIFT

The educator Mike Caulfield turned these habits into a simple method called SIFT:

1. Stop. Before you read, share or react, pause. Ask whether you know the source and whether the claim is something you need to check. Notice if it is making you emotional.
2. Investigate the source. Spend a minute finding out who is behind it and what their expertise and agenda are. Lateral reading is the main tool. A reference site like Wikipedia can be a quick starting point.
3. Find better coverage. Look for other reliable sources reporting the same claim. If trusted outlets or experts are not reporting it, be cautious. Often you can skip the original source and find a better one.
4. Trace claims, quotes and media to the original context. Follow a claim back to where it started: the original study, full speech or first appearance of a photo. Context is often lost or changed along the way.

The method is designed to be fast. Most checks take only a minute or two.

Reverse image search

Many misleading posts use real images in a false context. A reverse image search lets you upload an image, or paste its address, into a search tool to find where else it has appeared online. If a photo said to show yesterday's event turns up in articles from years ago, you have found a false context. For videos, you can capture a still frame and search that. Pay attention to the earliest version you can find and to its original caption.

Checking statistics

Numbers can sound authoritative while misleading. Useful questions include:

• Where does the number come from? A named study or official dataset, or nowhere at all?
• Relative or absolute risk? “Doubles your risk” sounds alarming, but a rise from 1 in 10,000 to 2 in 10,000 is a small absolute change.
• Compared to what? Large totals may simply reflect large populations; per-person rates are often more meaningful.
• Correlation or causation? Two things rising together does not mean one causes the other.
• Who was measured? A small or unrepresentative sample, such as an online poll anyone can join, cannot speak for a whole population.
• Which time period? Choosing a convenient start or end date can make a trend look rising or falling.
• How is the chart drawn? A vertical axis that does not start at zero can make small differences look huge.

Worked example: the coffee claim

A post with a striking graphic says: “New study: coffee drinkers are 50% more likely to develop a rare eye condition!” It links to a site called the Health Truth Report. Apply SIFT.

Stop: You drink coffee, and the claim makes you anxious. That emotional reaction is a signal to slow down, not to share.

Investigate the source: You open a new tab and search for “Health Truth Report.” Within a minute you find that it is run by a company selling dietary supplements, including a “coffee replacement” drink. That does not prove the claim false, but it is a clear conflict of interest.

Find better coverage: You search for news about coffee and the eye condition. A health news site and a university press release both describe the study, with more detail and less alarm.

Trace the claim: The press release links to the original study. You learn:

• The condition affects about 2 in 10,000 people. Among heavy coffee drinkers, the rate was about 3 in 10,000. That is a 50 percent relative increase, but an absolute increase of only 1 person in 10,000.
• The study was observational. It found a correlation, and the authors note that coffee drinkers in their sample also smoked more, which could explain the difference.
• The researchers call for further study and do not recommend that anyone stop drinking coffee.

The original post was not entirely invented, but it stripped away context, presented relative risk without absolute risk, and turned a correlation into an implied cause. The whole check took about five minutes. Many claims can be resolved even faster.