Corrections Are the Only Honest Metric

Trust in the press is measured constantly and almost always through surveys, which capture how people feel about a category rather than how a particular outlet performs.

There is a better measure and it is sitting in plain sight. An outlet's correction record shows how often it gets things wrong, how quickly it acknowledges it, and whether it treats a correction as an obligation or an embarrassment to be minimized.

The comparison is unflattering in a specific way. Outlets that publish many corrections are usually the ones with the most rigorous checking, because a process that catches errors after publication is generally attached to one that catches most of them before. An outlet with no corrections is either extraordinary or is not looking.

The practice worth demanding is simple: corrections on the page, dated, describing what was wrong rather than that something was, and never a silent edit. It costs nothing except the willingness to be seen being wrong, which is exactly why it is a real signal.

The survey approach fails for a reason worth stating plainly. It measures a feeling about a category, and the feeling tracks partisanship far more closely than it tracks anything a newsroom does.

An outlet cannot move that number by being accurate, which means the metric gives no guidance to anyone trying to improve. A measure that cannot be acted on is not a measure of performance.

The correction record has the opposite property. It is generated by the outlet's own process, it is dated, and it is specific enough that a reader can check whether the thing corrected was material.

The counterintuitive part is the direction. More corrections usually indicates better checking, because an organisation that looks for errors finds them and one that does not look publishes clean.

That inversion is why the metric is never used in marketing, and it is exactly why it is trustworthy. Nobody has an incentive to inflate it.

Government publishing solved the mechanics of this a long time ago and the model is instructive. The official gazette issues corrections as dated documents in the same series as the original.

The correction does not replace the error. Both remain, in sequence, so the record shows what was published, when it was wrong, and when it was fixed.

The permanent version preserves that sequence indefinitely, which means a reader in ten years can reconstruct exactly what a reader at the time would have seen.

The contrast with the silent edit is the whole argument. An article quietly amended presents itself as always having been correct, which is a small untruth told to every subsequent reader.

The codified text handles the same problem the other way, showing the current state while preserving the amendment history that produced it. Both approaches are honest. Neither pretends the earlier version did not exist.

None of this is expensive. A dated correction note costs a paragraph, and the records obligations that public bodies already operate under show that preserving supersession is a solved problem at far greater scale.

What it costs is the appearance of never having been wrong, which is the thing the practice is actually trading. An outlet that publishes its corrections is making a claim about its process rather than about its infallibility, and only one of those is checkable.

The objection is that a correction count rewards volume, since an outlet publishing more will correct more. That is right and it is why the useful comparison is proportional and qualitative rather than a raw total.

What a reader can assess without any arithmetic is the character of the corrections. An outlet correcting names and dates is checking. An outlet correcting the substance of its claims repeatedly is doing something else, and both are visible on the page.