Trust is measured as a number and produced by mechanisms nobody counts.

That confidence in news organisations has fallen to historic lows, that the gap between political camps has widened, and that younger audiences take their news from social platforms while trusting those platforms unevenly. Long-run series from large survey institutes carry this, with the usual sampling error attached to small year-on-year moves.

The finding that committed partisans on both sides read the same balanced report as biased against their own camp. It is the most robust result in this field and the least comfortable one: a piece can be fair and be received as an attack by everybody involved.

The mechanism. A number says how much trust exists; it does not say what produced it, what would change it, or why a correction sometimes strengthens the belief it was meant to fix. The sections below name the mechanisms that psychology has described and that communication reporting rarely mentions.

Reasoning that serves a conclusion the person already prefers. Evidence that fits is examined lightly, evidence that threatens is examined hard. It explains the common failure of correction campaigns: supplying more accurate facts to somebody who has a reason to reject them supplies more material to reject.

The choice of sources that already agree. It matters for coverage because it decides who ever sees a piece. A placement in an outlet your critics avoid reaches the people who were persuaded already.

Repetition makes a statement feel true, whatever its source. For an organisation this cuts both ways: a false claim repeated across three small outlets can outweigh one correction in a large one, and a true statement repeated in your own words gains the same familiarity.

“tendency to believe false information to be correct after repeated exposure”

Reference: Wikipedia, Illusory truth effect. https://en.wikipedia.org/wiki/Illusory_truth_effect

The delayed influence of a low-credibility source. Over time people remember the claim and forget where it came from, so a dismissible piece can become an undisputed impression weeks later. That is why a weak source is worth answering, even when nobody took it seriously at the time.

Source memory is the recall of where a statement came from. It is weaker than memory for the statement itself, which is the quiet problem with being quoted: the sentence travels, the attribution falls away. In generated answers the same thing happens mechanically.

The push-back that follows a message experienced as pressure. Campaigns written to persuade can produce the opposite of persuasion in an audience that already suspects motives. The practical consequence is a change of form: evidence that a reader can check produces less reactance than an appeal that asks to be believed.

Evaluating information by what accepting it would cost inside one's own group. It looks like a factual disagreement and behaves like a loyalty test. Where it is at work, better data changes nothing, and a shift comes only from somebody the group already trusts.

The assumption that one sees the world as it is, and that people who disagree must be biased or misinformed. It is the reason two camps can each cite the same article as proof of the other side's bias.

The belief that media affect other people more than oneself. It shapes what audiences demand: measures aimed at protecting others, and resistance to anything that treats oneself as susceptible.

“tend to perceive that mass media messages have a greater effect on others than on themselves”

Reference: Wikipedia, Third-person effect. https://en.wikipedia.org/wiki/Third-person_effect

Hostility towards the other camp as people, apart from any disagreement about policy. Where it is high, an outlet is judged by who else reads it, and a placement can damage standing with one audience precisely because it persuaded another.

Beliefs survive the removal of the evidence that produced them. A correction reaches the record and rarely reaches the impression. The realistic goal of a correction is the archive and the next reader, rather than the mind that already decided.

The deliberate turning away from news, often for reasons of mood and overload rather than disbelief. It is routinely counted as distrust, and the two require opposite responses: distrust asks for evidence, avoidance asks for relevance and brevity.

The ordinary human capacity to filter what is heard: who says it, what interest they have, whether it coheres with what is known. Scepticism read through this lens is a working faculty rather than a defect, and material that survives it is material that names sources and admits limits.

Confidence in institutions moves together: press, courts, administration, science. A decline in one is rarely explained by the conduct of that one. For an organisation this means the ground can shift without any change in what it does or how it is covered.

The competence to judge sources, claims and interests. It is the audience-side answer, and it is the one that outlasts a campaign. Material that teaches a reader to check is the only kind that keeps working when the reader no longer trusts anyone, including you.

Five things. Repetition beats volume. A weak source is worth an answer. A correction serves the record rather than the impression. Persuasion in the wrong form produces resistance. And material that lets a reader check something survives the loss of trust in everybody, which is the condition the surveys actually describe.

That a judgement can travel as a fact. A passage from an opinion piece carries the writer's position and enters an answer without it. Source memory fails in the same direction, only faster, because the system separates claim and attribution by design. The check is the same as for a human reader: does the sentence name what it rests on?