Study reveals positive impact of generative ghost AI chatbots

Sixteen bereaved individuals reported positive experiences after communicating via online video sessions with generative ghost AI chatbots programmed to simulate their deceased relatives. The academic project, which evaluated the psychological impact of digital replicas, recorded a unanimous desire among participants to repeat the interaction.

The University of Colorado Boulder published the findings in the Proceedings of the 2026 Designing Interactive Systems Conference. Information science PhD candidate Jack Manning led the research alongside associate professor Jed Brubaker. The academic team developed two distinct chatbot variations for the experiment, comparing a first-person simulation, termed reincarnation, against a third-person model, termed representation.

Participants consistently preferred interacting with the first-person generative ghost AI chatbots. One 32-year-old female participant, who communicated with a digital replica of her late grandmother, described the texting experience as visceral. She explained that she could vividly feel and see her relative, adding that the interaction provided her with long-awaited emotional closure.

The study revealed that participants tolerated factual errors from the software but reacted strongly against unnatural language choices. For example, a chatbot simulating a participant's stepfather incorrectly used the nickname "champ," which nearly compromised the experiment. Users favoured brief, emoji-rich replies over lengthy text blocks, indicating that emotional authenticity relied heavily on conversational rhythm and specific word choices.

Commercial enterprises like Project December and Séance AI currently market similar services compiled from the diaries, social media posts, and personal messages of deceased individuals. Brubaker explained that the rapid commercial advancement of the technology necessitated immediate research into its psychological effects. Despite the positive feedback from the study, Manning expressed significant caution regarding the unregulated deployment of generative ghost AI chatbots. He warned against the unrestricted use of these automated simulations, particularly for vulnerable individuals dealing with recent grief.