When an AI-generated diagnosis begins to matter: the AI consequence point

A patient can arrive at a consultation already organised around an AI-generated disease name. When does that use become relevant to diagnostic quality and safety - not merely to AI adoption? We define the AI consequence point: the first identifiable moment at which a patient's previsit use of diagnostic AI materially changes what happens next in care. This change may involve whether or when the patient seeks care, how the problem is presented, which explanations are considered credible, or how diagnostic decisions are negotiated. The concept refers not to AI use itself, but to the point at which that use becomes consequential for the diagnostic process. Curiosity, momentary surprise, and a label that leaves no durable trace do not qualify. AI exposure is not the AI consequence point; an AI label is not a diagnosis. Benefits and harms are both possible: earlier appropriate care and better preparation, or delay, fixation, narrowed attention, and authority conflict. Diagnostic safety research should measure previsit framing, prespecify observable material changes in the three domains, and examine how clinical responses amplify or correct them.