How to read the news critically: a field guide

News reporting tells you what happened, based on what the reporter has seen, documents and what people have said. Analysis explains what it means. Opinion argues for a view. They are all valuable, and they should not be confused: an editorial is a case, not a record. Sponsored content looks like a news article and is paid for by an advertiser, and is normally labelled as such. Much of the copy in different outlets also comes from the same wire service, such as PTI, ANI, Reuters or AP, so seeing the same story in ten places does not mean ten sources confirmed it.

Old stories are shared again and again as if they were new, and photos from one event are posted as photos of another. Before you share, check the date, and find where the claim started. A report that says officials said or a study found should lead you to the statement, the document or the paper. If you cannot find it, say so before you repeat it.

Headlines are written to attract a click and to fit a space, and they often say more than the article does. A headline that says a drug cures a disease may sit over a report on a small study in mice. Read the opening paragraphs, and look for the qualifiers, such as may, could, alleged and according to, which carry the real strength of the claim.

Numbers carry authority and are easy to misuse. A few checks:

• Compared with what? A figure with nothing to compare it to, a rise of 40 per cent from what, tells you little.
• Absolute or relative? A risk that doubles from 1 in a million to 2 in a million is still very small.
• What is the denominator? Totals and rates can give opposite impressions: a big state has more cases than a small one without having a worse problem.
• Does the chart start at zero? A bar chart that starts partway up makes a small difference look large.
• How was it measured, by whom and when?

Researchers who study how people judge online information found that professional fact-checkers do something that most readers do not. When they meet an unfamiliar site, they do not read it closely. They leave the page and open new tabs to see what other sources say about who is behind it. This is called lateral reading, and it is quick and far more effective than reading the site's own About page. Do the same with a social media account that is making a surprising claim.

A picture can be real and still be from another time or place. A reverse image search, available in most browsers, shows where else it has appeared. Video can be edited or taken out of context, and altered or generated images are increasingly good. Messages on private apps that say forwarded many times, or tell you to send them to everyone you know, are a warning in themselves. Fact-checking organisations in India and elsewhere publish what they have checked, and the government has its own fact-check unit; no single checker is perfect, so look for a conclusion that several agree on.

The first reports of a fast-moving event are often wrong in details, which are corrected later, so wait for updates before settling on a view. Look for a corrections policy: the reputable outlets publish corrections and say what changed. And check yourself too. We are all more ready to believe what we want to be true, as our guide to cognitive biases describes, and the best protection is to read sources whose views differ from your own.

Plax summaries are meant as a starting point. Each one links to the original story at its publisher, and the headline is the link, so that you can read the full report whenever something matters to you.

• Cognitive biases that shape how we read the news

  Confirmation bias, the availability heuristic, framing and more: how our minds distort what we read, and what helps. A guide for readers of the news.

• How science works: hypotheses, evidence and peer review

  A short guide to how scientific claims are tested and checked, and why a headline that says a study shows something deserves a closer look.

• How AI chatbots and summaries work, and why they sometimes get things wrong

  Large language models predict text. That makes them fluent and useful for summaries, and it also explains why they can state falsehoods with confidence.