Judging Online Sources

Almost everything you read in English now arrives without the context a newspaper page used to provide. Working out what kind of thing you are reading is now the reader's job.

What kind of page is this?

News report — reports events, attributes claims, names sources Opinion / comment — argues a position; usually labelled, sometimes only in small print Sponsored content — an advertisement written to look like an article. Look for sponsored, promoted, partner content, in association with Press release — written by the organisation it is about; no independent verification Aggregator — repackages someone else's reporting, often with a stronger headline

Sponsored content is designed to be mistaken for reporting. The label is legally required but deliberately small.

The address bar tells you something

Read the domain from the right, not the left. In news.example.co.uk/health, the site is example.co.uk — everything before it is theirs to invent.

• example.com.news-today.net is not example.com. The real domain is the last part.
• .gov and .ac.uk / .edu are restricted and reasonably reliable signals of an official or academic source.
• .org is not restricted — anyone can register one, including campaign groups and companies.
• A long string of hyphens and words is a mild warning sign on its own.

Questions worth asking

Who wrote it? A named author with a track record is different from "Staff Writer" or nobody.

Who pays for the site? Advertising, subscriptions, an industry body and a campaign group all produce different incentives.

When was it published? An undated page is a warning. A five-year-old statistic presented as current is a common problem.

What is the source of the numbers? "Studies show" with no link means nothing. A link to the actual study lets you check whether it says what is claimed.

Does the headline match the article? Headlines are often written by someone else and overstate the content. If the body hedges where the headline asserts, trust the body.

Read sideways, not down

The instinct is to evaluate a page by studying the page. That is backwards, because a page can say anything about itself.

Lateral reading means leaving it: open a new tab and search for the site's name, the author, or the claim itself. Thirty seconds elsewhere tells you more about who is behind a page than thirty minutes of reading it carefully — and it is what professional fact-checkers actually do.

Screenshots and quotes

Material reaches you stripped of its context more often than not:

• A screenshot of a post has no date, no thread and no proof it is real.
• A quote with no link is unverifiable — and quoting half a sentence is the standard way to reverse someone's meaning.
• Old stories recirculate. Check the date before reacting; a genuine article about a real event can be five years old.
• Outrage travels faster than corrections. If something is designed to make you angry enough to share it immediately, that is a reason to wait.

Machine-written pages

A growing share of what ranks well is generated to fill space rather than to inform. Common signals: no named author, a date but no update history, confident claims with no sources, oddly even coverage of a topic with nothing specific in it, and text that never says anything a person would have had to know.

The test is the same as always: does it cite anything checkable? Generated text is fluent; what it usually lacks is a verifiable source.

The pattern to watch for

Headline: Coffee cures anxiety, study finds

Body: …a small observational study found participants who drank coffee reported slightly lower anxiety scores. Researchers cautioned that the effect may reflect other lifestyle factors.

Every step downgrades the claim. This is extremely common in health and science coverage, and the headline is the part that gets shared.

Signals of persuasion rather than information

• Emotional language in a supposedly factual piece
• Only one side quoted
• Statistics with no source
• "Experts say" without naming any
• Urgency: act now, before it's too late
• An unrelated conclusion: an article about a problem that ends by recommending one product

Check hardest what you already agree with. Everyone applies scrutiny to claims that contradict them; almost nobody applies it to claims that confirm them. A page that tells you what you already believe, from a source you have never heard of, with no citations, is exactly the one most likely to get shared unchecked — and the one most worth a thirty-second search.

Practical habit

Before sharing or citing anything: read past the headline, check the date, and find the original source. Three checks, under a minute, and they catch most of it.

Test yourself

The quiz below asks what kind of page you are looking at.