Aims: Recent advances in generative artificial intelligence (AI) have enabled “deathbots,” interactive simulations of deceased individuals intended to sustain continuing bonds and provide emotional comfort. Their clinical role remains uncertain. We synthesised their clinical, ethical and social implications for oncology and developed a preliminary framework for responsible integration.
Methods: PubMed was searched on 22 July 2026 for postmortem AI literature, yielding 615 records; 563 were published in 2015–2026. An effects-focused search identified 38 records, all of which were included following full-text assessment. Additional literature identified through citation tracking and hand searching was also included. A web-based market scan characterised commercial services and stated limitations. Findings were synthesised thematically.
Results: The literature was sparse and heterogeneous, dominated by conceptual analyses, hypothetical scenarios and attitude studies. Direct evidence from bereaved users comprised one qualitative study of three cases. Adjacent non-generative digital legacy research in oncology and palliative care suggested feasibility and potential benefits for dignity, meaning, communication and continuing bonds, but quantitative findings were inconsistent and not transferable to postmortem AI. Anticipated benefits—comfort, unfinished communication and memory preservation—were counterbalanced by dependency, disrupted grief adaptation, identity distortion, privacy, family conflict, commercial exploitation and service discontinuation. Commercial services varied in modality, fidelity and provider-determined pricing, without standardised quality benchmarks or assured continuity. We derived a five-domain framework: prospective consent and preferences; authenticity and transparency; recipient and family agreement; clinical risk assessment and human support; and data, commercial and lifecycle governance.
Conclusions: Postmortem AI may offer therapeutic potential and is commercially available, but evidence for routine clinical implementation remains limited. Incorporating postmortem AI discussions into oncology advance care planning is increasingly relevant, creating an urgent need for governance addressing ethical concerns and psychological burden. The framework requires empirical validation and longitudinal safety evaluation.