The 7 dimensions of journalism quality
Why we separate accuracy, evidence, sourcing, fairness, context, clarity, and accountability — and what each one actually measures.
“Quality” is one of those words that survives meetings because everyone assumes they mean the same thing. They usually do not. For one editor it means accuracy. For another it means prose. For a reader it often means “did this help me understand what happened and whether I can trust it?”
MediaMeter separates quality into seven dimensions because a single score hides the disagreements that matter. An article can be accurate and unfair, clear and thinly sourced, well-contextualised and impossible to challenge. Collapsing those into one number does not resolve the tension; it conceals it.
1. Accuracy
Accuracy asks whether claims match the best available evidence at the time of assessment. It is claim-level, not article-level. A piece with nine correct claims and one wrong one is not “90% accurate”; it contains a material error.
Useful criteria include whether numbers, names, and dates are correct; whether a correction history exists; and whether the headline asserts something the body does not support. Accuracy is not the same as certainty. A carefully hedged claim can be accurate; a confident claim can be wrong.
2. Evidence
Evidence asks whether key assertions are traceable to named, checkable sources. This is different from accuracy. A claim can be true and unsupported in the text. A claim can cite a source that exists but does not say what the article implies.
We look at source proximity — is the source primary or repeating someone else? — and at the document trail. “According to reports” is weaker than a link to the report. On-record attribution is stronger than an unnamed official, though anonymity is sometimes necessary and should be explained.
3. Sourcing
Sourcing is about range. Who is heard, who is missing, and whether the mix fits the story? A labour dispute covered only through the employer’s statement is not balanced, even if every sentence is factually correct.
Criteria include source diversity, the balance of primary and secondary sources, and disclosure of conflicts. Sourcing is context-dependent: a court report does not need the same range as a policy analysis. That is why weights vary by content type.
4. Fairness
Fairness asks whether competing positions are represented without loaded framing. It is not the same as false balance. Giving equal weight to a fringe claim and a scientific consensus can be unfair to the reader.
We look for right of reply, balanced characterisation, and whether quotations are selective in a way that changes meaning. Fairness also includes describing people as they describe themselves, where relevant.
5. Context
Context asks whether readers get the background needed to interpret the claim correctly. A accurate number without a base rate can mislead. A crime statistic without a trend can imply a crisis that is not there.
Criteria include temporal context, base rates, and known caveats. Context is often the first casualty of breaking news, and the dimension readers notice most when it is missing.
6. Clarity
Clarity asks whether the language is precise enough that reasonable readers reach the same meaning. This is not about short sentences or plain style alone. It is about whether the headline matches the body, whether terms are defined, and whether numbers imply precision the data does not support.
“Up to 40%” is clear. “40%” may not be. “Linked to” is not “caused by.” Clarity failures often look like accuracy failures, which is why they need separate measurement.
7. Accountability
Accountability asks whether the work shows who is responsible and how to challenge it. That means a byline, a dateline, a corrections route, and a contact. It also means being explicit about what the newsroom does not know.
An article can be accurate and still unaccountable if it hides its author, its funding, or its correction policy. Accountability is the dimension most often missing from automated quality tools because it is structural rather than textual.
Why seven, and not one
One score is useful for a dashboard and useless for a decision. If an article scores 62, what should an editor do? If it scores 88 on accuracy, 41 on evidence, and 79 on clarity, the action is clearer: find the missing sources, keep the prose.
Seven dimensions also make disagreement visible. Two editors can agree an article is “good” and disagree about why. That disagreement is useful; it is how standards evolve.
Why these seven
They map to established codes — SPJ, IFJ, AFP, IFCN — without pretending those codes are interchangeable. They cover the questions readers actually ask: Is it true? Can I check it? Is it fair? Do I understand it? Can I challenge it?
They are not exhaustive. News judgment, originality, and public interest matter, but they are properties of story selection rather than article quality. We keep the framework focused on what can be assessed with evidence.
What the framework does not do
It does not determine intent, legal liability, or whether a story should have been published. It does not rank outlets. It does not replace an editor. It produces structured findings that a person can inspect, contest, and improve.
That is the modest claim. A quality framework is useful precisely because it is modest: it makes the conversation specific.