Retractions are rare and heavily covered. Corrections are frequent, small and mostly unread, and they are the better indicator of whether an outlet can be relied on.
Most errors are ordinary rather than scandalous
The typical mistake is a transposed figure, a misattributed quotation, a wrong job title or an outdated fact carried forward from earlier coverage.
None of these produce a scandal, and most readers never notice them, which means correcting them depends entirely on the outlet's own practice.
An organisation that corrects such things promptly is one that treats accuracy as routine maintenance rather than as reputation management.
The mechanics reveal the priority
A correction can be appended at the foot of an article, placed prominently at the top, or made silently by editing the text with no notice at all.
Silent editing is the most damaging option, because a reader who saw the original has no way of knowing the record has changed.
Outlets serious about the practice publish a dated note explaining what was wrong and what it now says, which allows anyone to reconstruct the history.
Reachability is the limiting factor
An error can only be fixed if someone can report it, and the accessibility of that route varies enormously between publications.
Where the only channel is a general contact form, reports from readers are effectively lost, whatever the stated policy says.
A named corrections contact and a published policy are cheap to provide, and their absence is informative about how often the organisation expects to use them.
Distribution outlives the original
Once an article has been aggregated, quoted and shared, correcting the source does not reach the audiences who encountered the copies.
Some outlets notify syndication partners when a correction is issued, which limits the spread. Most do not, and the erroneous version persists indefinitely.
Reading the record is straightforward
Corrections pages are public, and reviewing a month of them shows how many errors an outlet catches, how quickly, and whether the notes are candid or evasive.
A page with almost nothing on it is not evidence of accuracy. It usually indicates that errors are handled quietly or that nobody is checking.
This is one of the few quality signals a reader can verify directly without any specialist knowledge of the subject being reported.