From fact verification to understanding misleadingness: a survey and roadmap on reader-centric multimodal misinformation detection

Misinformation spreads widely not only because of factual inaccuracies but because it evokes strong emotions, activates moral engagement, and is weakly anchored in context, allowing different audiences to derive divergent interpretations from the same content. This interpretive variability highlights the need to move beyond fact verification towards a deeper understanding of misleadingness. Drawing on insights from psychology, linguistics, media theory, and communication studies, we identify three core dimensions that shape misleadingness: emotion, communicative intent, and context. Building on these dimensions, we reconceptualise multimodal misinformation detection from the reader's perspective and propose a Reader-Centric Multimodal Misinformation Detection framework. We then use this framework to survey existing literature and to examine whether existing datasets, models, and evaluation practices adequately support reader-aware multimodal learning. By reframing misinformation detection as a problem of interpretive meaning-making, this work bridges computational approaches with human understanding and outlines key directions for building truly reader-aware misinformation detection systems.