Study Reveals AI Can Offer Comfort Through Conversations with Deceased Loved Ones

A recent study published in the Proceedings of the 2026 ACM Designing Interactive Systems Conference documents how individuals can engage in simulated conversations with deceased relatives using advanced AI models. The research, led by Jack Manning, a PhD candidate at the University of Colorado Boulder, involved sixteen participants who interacted with AI representations of their late loved ones through various digital platforms. One participant reported feeling a profound sense of closure after conversing with an AI-generated version of her grandmother.

The study utilized large language models to generate these "generative ghosts," which simulate conversations based on data such as personal writings, voice recordings, and photographs of the deceased. Participants engaged in approximately twenty-minute sessions where they interacted with two versions of the AI: one speaking directly as the deceased person and another describing them in the third person. The facilitator guided the conversation while a researcher inputted information into the model to create a realistic interaction.

Surprisingly, participants generally found these interactions positive rather than unsettling or eerie. Manning noted that he initially expected people would find such conversations uncomfortable but was proven wrong by their overwhelmingly positive reactions. This suggests that AI-generated conversations with deceased loved ones could offer significant emotional support and closure for those grieving the loss of family members or friends.

The implications of this research extend beyond personal grief management, touching on broader questions about how technology can address human needs in innovative ways. As these technologies evolve, they may provide new avenues for healing and connection, potentially transforming how society views and copes with loss.