Reading character, not just relevance: Query, person, and content as a three way model of search preference, and the discovery that personality has a felt layer and an enacted layer

Search ranking optimises relevance and authority: how well a result matches the query, and how much standing it holds. Neither captures the character of the content, how a page reads (rigorous or warm, original or derivative, selling or explaining) as opposed to what it is about. We ask whether that character, matched to the character of the searcher, their disposition, changes which result a person prefers, and under what conditions. We model search preference as a three way interaction between the query phrasing, the searcher's personality (DYNAMICS-8, an eight trait model), and the content's character (DYNAMICS-WEB, an eight axis measure of how content reads). The work has an honest arc. Naive tests found nothing; we trace the null to three specific faults and fix each. With the instrument repaired, the query emerges as the dominant signal, and personality as a conditional modifier that acts only when the query is ambiguous. A single topic then suggested a rich one trait per axis personalisation matrix, which a fair multi topic replication demolished, leaving a single durable link: acuity predicts a preference for depth. We then leave simulation behind and measure a real person end to end: from six months of their browsing we score the pages they actually read, infer their personality blind from behaviour, and find it predicts their revealed content character on six of seven axes, with the surviving link holding against a psychometric questionnaire. Finally, measuring the same person three ways, a natural self report, a disciplined self report, and the personality read off their behaviour, uncovers a clean structure we did not expect: personality has a felt layer and an enacted layer. Cognitive traits are identical across all three measurements; self regulation traits form a gradient from disposition to behaviour. Behaviour reads the enacted layer, and the enacted layer is what predicts what a person consumes. We give the product implication (a layered ranker, and the case for behavioural rather than self reported profiles) and we are explicit about the limits: most evidence is a calibrated simulation, the human result is one person, and the path to a definitive claim runs through the same pipeline applied to many consenting users.