NEW YORK – The app costs $24 a month. For that subscription, a startup called Re;memory will build three animated chatbot avatars of people who have died, trained on text messages, emails and whatever voice recordings a family can provide, and let subscribers keep the conversation going indefinitely. The company markets the service as a grief tool. Researchers at the University of Colorado Boulder have a different term for it: a generative ghost.
Generative ghosts are the newest and most ethically contested corner of the AI industry. Three companies now compete openly in the space: Re;memory, which charges $24 monthly; Séance AI, which offers animated image interactions for $19.99 a month; and You Only Virtual, a third entrant offering similar capabilities with different packaging. All three say their user bases are growing. None will disclose subscriber numbers.
Jack Manning, a doctoral candidate studying technology and death at the University of Colorado Boulder, has tracked the sector’s emergence. “It is definitely a growing sector,” he told CBS MoneyWatch. “We’re seeing growth in interest in how AI and grief intertwine.” Manning and his colleague Jed Brubaker, an associate professor at the same institution, have spent years studying bereavement practices across digital platforms. Their framework puts generative ghosts in a category distinct from earlier forms of posthumous digital presence, like memorial Facebook pages or automated email services that send messages after a person dies.
These services work by ingesting the digital footprint of a life. A subscriber provides text message archives, email threads, social media posts, and voice recordings. The AI model analyzes linguistic patterns, response cadence, and vocabulary, then builds a conversational agent that generates responses in the style of the deceased. The result is not a replay of something the person actually said. It is a probabilistic prediction, generated in real time, of what that person might have said given the prompt.
One legal and ethical question the sector must answer is whether these services constitute deepfakes. Manning draws a principled distinction. “The fundamental premise of a deep fake is the intent to deceive,” he said. “The intent of generative ghosts is not to deceive another person.” A person using Re;memory or Séance AI knows exactly what they are interacting with. That argument distinguishes these products from synthetic media designed to impersonate someone without their audience’s knowledge, which has driven legislation in several US states and the European Union.
The rapid commercial expansion of AI services into grief and bereavement follows a broader pattern in the technology industry. After the broader commercial AI race produced clear winners in monetization-focused platforms, entrepreneurs have sought applications in emotionally charged domains that earlier generations of software left largely untouched. Death and grief, with their associated industries valued in the hundreds of billions globally, were an obvious frontier.
The subscription model signals commercial intent, but it also creates a recurring financial relationship between a company and a bereaved customer that critics find unsettling. Re;memory’s $24 tier includes three avatars, meaning a single subscriber could maintain chatbot versions of a parent, a sibling and a spouse simultaneously. The pricing, low enough to persist quietly on a credit card statement, is designed for long-term retention.
What is missing from every company’s marketing materials is longitudinal evidence of what the experience does to grief over time. Manning and Brubaker say the sector is too new to have produced it. Does sustaining a chatbot of a deceased parent help the bereaved process loss, or does it forestall that process? The current answer is that no one knows, because the services have not existed long enough to find out.
The services also raise questions about data consent that extend beyond the subscriber. The deceased person whose messages are being used to train the model never agreed to that use. In most cases, there is no legal framework that explicitly governs posthumous data licensing. The companies in this space operate in a regulatory gray zone that neither the Federal Trade Commission nor Congress has addressed directly, leaving the ethical framework to researchers and the market. Major AI platforms, meanwhile, have kept their distance from the grief sector entirely.
Manning frames the entire sector as part of a question the technology industry has not finished asking: what happens to the digital representation of a person after they die? Social media platforms have accumulated billions of accounts belonging to people who no longer exist. AI companies have built models on data that includes the writing, images, and recordings of the dead. Generative ghost startups are the logical product of that accumulation. Whether they are also a comfort or a complication for the people who pay for them remains, for now, an open question that grief counselors and computer scientists are trying to answer at the same time.
