In intimate messaging, how a difficult note is produced signals effort and ownership. We study how AI assistance level (light tone rewrite vs heavy full draft) and a brief co-sign disclosure shape receiver attributions and outcomes in apology and boundary setting. Study 1 (N=152) instrumented authoring to build a many-stimuli corpus. Study 2 (N=704) tested effects in a mixed effects experiment. Heavier drafting reliably reduced perceived ownership and authenticity; clarity/competence gains did not compensate. In apologies, co-signing a tone rewrite increased authenticity and forgiveness; co-signing a full draft decreased both. In boundary requests, co-signing yielded small or negative shifts. Stimulus-level analyses tied idiosyncratic "voice" cues to ownership/authenticity, and revealed sender-receiver miscalibration. We contribute: (i) scenario-aware causal estimates for help and disclosure; (ii) an empirically grounded, scenario-aware attributional account alongside a competence–integrity dissociation; (iii) evidence of sender–receiver miscalibration; and (iv) design guidance voice preserving defaults, ownership restoring scaffolds, and CPM disclosure.
ACM CHI Conference on Human Factors in Computing Systems