Synthetic corroboration: when AI interpretation looks like independent evidence

People increasingly use generative artificial intelligence (AI) to interpret the motives, emotions, and likely actions of absent others. This article develops synthetic corroboration, defined as the inflation of perceived independent target-specific evidential support when an externally attributed AI interpretation remains substantially dependent on evidence supplied by the user. The construct distinguishes independence of target-specific source access and information from analytical independence: AI may contribute general knowledge, alternative hypotheses, or inferential processing without providing an independent evidential source. Drawing on research on interpersonal inference, advice taking, source dependence, judgement aggregation, source monitoring, machine agency, and social sycophancy, the article proposes a programme of discriminating tests. The core experiment crosses source attribution with structural transformation while holding target-specific evidence constant, comparing AI with self-reformulation and a human adviser given the same evidence. The central prediction is that transformation will disproportionately increase perceived evidential independence under AI attribution and reduce subsequent acquisition of genuinely independent information. AI–human equivalence, failure to predict information seeking, or reduction to established constructs would narrow or defeat the AI-specific account.n psychological fluency or agreement.