How a story was sourced, framed and funded, and what got left out.
Most arguments about bias are really arguments about who was asked. Start with the source list instead of the headline and the whole dispute changes shape.
Open any article, then set each criterion to what you found on the page. The dial does the arithmetic with the same weights we use. It tells you where to look harder, not what to believe.
Worked explainers that show each check on an ordinary story, so the technique is visible without the politics.
SourcingA story can quote widely and still rest on nobody with first-hand knowledge. We separate witnesses from commentators, named sources from attributed ones, and documents from people describing documents, then show how a piece reads once you mark each quote with what the speaker could personally know.
FramingFraming rarely requires a false sentence. It requires a choice of baseline, a choice of timeframe and a choice of which comparison goes first. A walk through the standard moves, using an ordinary story with no villain in it, so the technique is visible without the politics.
Fact checksEach retelling drops a qualifier. By the third version the sample size, the confidence interval and the word associated have all gone, and a tentative finding is reported as a cause. We follow one claim down the chain and mark exactly where each caveat fell off.
CorrectionsLook for whether corrections are attached to the original article or buried on a separate page, whether they say what changed, and whether the outlet ever corrects itself unprompted. A publication's error handling predicts its reliability better than its tone does.
FundingCommissioned research is not automatically wrong, but who paid, what was asked and what went unreported all belong in the story. How to find the funder, how to spot the questions that were designed to produce a headline, and what a responsible disclosure line looks like.
NumbersA rise from two to three is one percentage point and a fifty per cent increase, and reporters choose between them for effect. The same choice governs rates versus raw counts and averages versus medians. A short field guide to the substitutions that make small changes sound enormous.
Every analysis answers the same fixed set of questions in the same order. Here is the whole set, with the weight each carries and what strong and weak work looks like.
Who was asked, and what could each of them personally know? Mark each quote with what the speaker could personally know, then reread the piece.
Which baseline, which timeframe, and which comparison went first? Ask what the same facts look like against the long run average, not the most recent low point.
Are the figures sourced, compared fairly and stated in the right units? Convert every percentage change into the raw numbers behind it and see if the claim survives.
Who paid for the research being quoted, and what was asked? Find who commissioned the research and read the questions they actually asked.
Does the headline claim what the body actually supports? Read the last three paragraphs first and check that they agree with the headline.
How does the publication handle being wrong? Check whether corrections are attached to the original article and say what changed.
We do not rate outlets on a left to right scale, and we do not grade mastheads. What we can say is where each kind of story usually needs the closest look. Turn a card for the checks.
Format 01Report
Fast, short, written to be updated, and often the first version anyone reads.
Common trapEarly copy is revised as facts firm up. Find the latest version before you judge the first.
Format 02Opinion
An argument from a named writer. Fine when labelled; a problem when dressed as a report.
Common trapA column laid out like a news report borrows a credibility it has not earned.
Format 03Report
A story built mostly from one organisation's announcement.
Common trapWhen the same quotes appear word for word across outlets, you are reading the release.
Format 04Numbers
A survey result turned into a headline about what people think.
Common trapA change smaller than the margin of error is reported as a shift in opinion.
Format 05Science
A research finding retold for a general reader, often via a press release.
Common trapEach retelling drops a qualifier, until a tentative finding is reported as a cause.
Format 06Visual
A picture or video moving faster than its context.
Common trapAn old image recirculates with a new caption and a new meaning.
Format 07Live
Rolling updates written while events are still unfolding.
Common trapAn early unconfirmed claim survives into the summary at the top.
Format 08Feature
Months of work, many sources, documents and a right of reply.
Common trapLength and confidence are mistaken for proof. Check the evidence for each claim.
Each analysis follows the same structure, so it can be handed to a class as a worked example. These are the questions in a form students can apply to a story they choose themselves.
0 of 5 questions asked
One story taken apart each week: the source list, the framing choice, the number that was doing the work, and the question the piece never asked. Short enough to read with a coffee, structured so you can run the same check on tomorrow's news yourself.
SubscribeNews Evaluation is about method, not verdicts. We take published journalism apart to show how it was built: who was sourced and what each source could personally know, how the framing was chosen, how numbers were presented, who funded the research being quoted, how the headline relates to the body, and what a reasonable reader would need that the story did not supply. We cover corrections culture as a beat in its own right, because how a publication handles being wrong is the most reliable public signal of how carefully it works. We do not rate outlets on a left to right scale. We evaluate individual pieces of work, including good ones, because the point is to teach a habit that transfers to whatever the reader opens next. One boundary is worth stating up front: we are not a news service and we publish no edition of our own. If you want the day's news reported and finished, that is our sister masthead The News Today. Our subject is the reading, not the day. How to read the news is the whole of our beat, and we do not stray outside it.
Every analysis answers the same fixed set of questions in the same order, so readers can apply the method themselves rather than trusting our conclusion. We link to and quote the original at sufficient length to be fair, and we read the whole piece before commenting on any part of it. We judge a story against what was knowable at the time it was published, not with hindsight. We say plainly when a piece is sound, and we distinguish an error of fact from a defensible editorial choice we happen to disagree with. Where we cannot establish something, we say we cannot, and we do not fill the gap with implication. News Evaluation is published by SUKH SANDHU PTY LTD (ACN 679 292 392, ABN 49 679 292 392) and is independent of the outlets it examines.
A site about other people's errors has to be exact about its own. Corrections are attached to the article that carried the mistake, dated, and written to say what was wrong and what it is now, never softened into a clarification when it was an error. If we mischaracterise a story or a journalist, the correction is as prominent as the original claim and we notify the publication concerned. We keep a running record of our corrections in the open, because a corrections page that stays empty is evidence of nothing except that nobody is looking. Corrections can be requested through the contact address on this site, and journalists whose work we analyse always get a reply.
No. Bias charts sort mastheads once and invite you to stop reading. We analyse individual articles, including strong work from outlets people distrust and weak work from outlets they like, because reliability varies story by story and the skill worth having is the one you can apply to a piece you have never seen before.
Yes, that is a large part of why it exists. Each analysis follows the same structure, so it can be handed to a class as a worked example, and the Classroom section carries the questions in a form students can apply to a story they choose themselves. Please credit the site if you reproduce material.
We do, through the contact address on this site. We choose based on whether a story illustrates a method that transfers, not on how strongly anyone feels about it, so we decline plenty of requests. Tell us which part of the piece troubled you; it helps more than telling us the conclusion you reached.
It is published by SUKH SANDHU PTY LTD, an Australian company, and it is not owned by, funded by or run in partnership with any of the news organisations whose work it examines. Any commercial arrangement that could bear on our analysis will be disclosed on the page it affects.