Jack Manning's personal experience with grief and memorialization has led him to explore how technology, particularly artificial intelligence (AI), can help families remember and interact with loved ones who have passed away. Over a decade ago, after losing his sister, Manning and his family devoted countless hours and efforts-holding bake sales and organizing 5K runs-to raise funds for a meaningful memorial project in her honor. This challenging and time-consuming process inspired Manning, now a doctoral candidate in information science at the University of Colorado Boulder, to consider whether AI could provide a new way to commemorate those who have died.
Manning is currently researching the use of large language models (LLMs), a form of AI that can process and generate human-like text, to develop what some refer to as "generative ghosts." These AI-driven chatbots are created by training on a deceased person's digital footprint, including social media posts, emails, audio and video recordings, photos, and other personal data. The resulting digital simulations enable family members and friends to interact with a virtual representation of their loved one after death, offering a novel way to maintain a connection.
This concept is part of a growing trend in the intersection of AI and grief. Several startups have emerged providing services that create digital avatars of deceased individuals, allowing users to engage with lifelike representations. Companies like Séance AI, You, Only Virtual, and re;memory offer platforms where users can upload photos, voice recordings, and other personal data to generate these avatars.
For instance, re;memory advertises its service as a way to find comfort by interacting with digital versions of lost loved ones. Their platform enables users to see and hear virtual avatars resembling their parents or family members as they appeared in treasured photographs, rather than relying solely on memories or imagination. This approach aims to provide solace and a means of expressing love and forgiveness, bridging the emotional gap left by a person's passing.
Academic research is also delving into how people respond to these AI-generated representations. In a June 2026 paper published by the Association for Computing Machinery, Jed Brubaker, an associate professor at the University of Colorado Boulder, along with Manning and colleagues, studied participants' reactions to AI "ghosts." Study participants were asked to complete surveys sharing positive memories, personal stories, and descriptions of the deceased's personality traits. The researchers noted that participants were intrigued by how much an LLM could do with relatively limited input data.
According to Manning, there are two main types of generative ghosts. The simpler form, sometimes called a "death bot," is limited to replaying recorded statements made by the deceased without the ability to simulate their unique communication style. These bots offer a basic digital memorial but lack conversational depth.
In contrast, more advanced generative ghosts utilize AI to recreate the deceased's voice, mannerisms, and conversational patterns. This allows for interactive experiences where users can ask questions and receive responses that resemble the way their loved ones might have answered when alive. Brubaker explains that just as one might prompt an AI to speak "like Shakespeare," it is possible to instruct it to "talk like my grandfather" using specific behavioral details. These models can even say things the person never explicitly expressed, effectively "talking for" the deceased in a familiar and recognizable manner.
The costs of accessing these AI memorial services vary. For example, re;memory charges $24 per month for users to create up to three custom avatars. Séance AI offers a subscription at $19.99 per month, allowing users to generate animated images of loved ones that can smile, move their heads gently, and speak using the deceased's voice. These offerings aim to balance affordability with the technological complexity required to produce convincing digital representations.
User preferences and emotional responses to these AI "ghosts" have been studied as well. The research indicates that most people prefer interacting with the chatbot in the first person, as if directly communicating with the deceased, rather than through an avatar speaking in the third person. Authenticity is crucial; if the AI uses terms of endearment or expressions that the loved one typically never used-such as "champ" or "pal"-users often feel discomfort and may choose to end the interaction. This highlights the importance of accurately capturing the deceased's personality and manner of speaking to create a meaningful and respectful experience.
While generative ghosts share some technological elements with deepfake media-manipulated videos or audio designed to deceive viewers-there is a fundamental ethical distinction. Brubaker emphasizes that the key difference lies in intent. Deepfakes are created to mislead or deceive the public, often for malicious purposes. In contrast, generative ghosts are designed with the intent of comfort and remembrance, not to deceive others but to provide a personal, digital form of mourning and connection.
As AI technology continues to evolve, these digital memorials may become an increasingly common way for people to cope with loss, offering new possibilities for preserving memories and maintaining bonds beyond death. Manning's work and the rise of related startups suggest that the fusion of AI and grief support is a growing field, one that combines cutting-edge technology with deeply human emotional needs. However, the development and use of these tools also raise important questions about authenticity, consent, and the ethical implications of digitally recreating individuals who have passed away.
In summary, the emergence of AI-powered generative ghosts represents a novel frontier in how society approaches grief and remembrance. By leveraging large language models and digital data, these virtual avatars offer a unique, interactive way to honor and engage with the memory of lost loved ones. Despite the technological promise, success depends heavily on faithfully capturing the individual's personality and respecting the emotional sensitivities of users. As this field grows, it will require ongoing research, thoughtful design, and ethical considerations to ensure that such digital memorials provide comfort without compromising authenticity or causing unintended harm.