Our phone rings, and when we answer, the voice on the other end is that of a deceased friend. We receive a birthday message from a deceased child. This is possible today, and it’s happening.
Thanabots , or grief robots, are artificial intelligence programs trained on information about a deceased person, obtained from their social media profiles and interactions, as well as other personal data. These traces of a human being’s life are transformed into a disturbing, speaking identity shaped by families, algorithms, and corporations.
What once belonged to the realm of séances, personal religiosity, or science fiction is gradually becoming a service for everyday consumption.
In 2013, an episode of the science fiction series Black Mirror titled “Be Right Back” presented the following scenario: the protagonist, Martha, loses her partner, Ash, in a car accident. So great is her grief that she uses a service that reconstructs Ash from his digital “digital footprint.” It begins with messages, then progresses to a familiar voice, and finally culminates in a synthetic body. This replica knows Ash’s typical phrases and jokes. While it can reproduce what he left behind, it cannot recapture what it meant to be him.
A similar case (without the synthetic body part) soon appeared in the real world. Eugenia Kuyda , co-founder of the chatbot Replika , used messages from her late friend, Roman Mazurenko, to create a conversational bot in his style.
Today, thanabots , or mourning robots, use messages, photographs, videos, voice recordings, and social media posts to simulate the speech and behavior of a deceased person. They are part of a growing digital “afterlife industry” that transforms traces of a life into interactive recreations.
Beyond experimentation
The concept has gone far beyond fringe experimentation. Microsoft holds a patent , granted in 2020, to “create a conversational chatbot of a specific person” using social data such as messages, voice data, images, and other personal material. The patent explicitly states that the “specific person” can be someone who has already passed away, whether a friend, a family member, a celebrity, or a historical figure.
The prices clearly illustrate the business logic. A recent report described how the app HereAfter AI (which announced its closure in the summer of 2026) offered plans for around US$199, while the South Korean company DeepBrain AI was said to offer high-end 3D avatar recreations for up to US$50,000 (not including maintenance fees).
From memory to speaking identity
People have always turned to objects to maintain a connection with the deceased. A coat can hold a scent. A recipe can evoke the voice of a loved one. A saved voicemail can condense years into seconds. But the oldest memorial objects convey something of the person without pretending to be them.
A thanabot changes that pattern. It speaks in the first person: “I remember,” “I miss you,” “I still believe,” even though these are new phrases generated after death. A photograph says: “This person was here.” A thanabot says: “I’m still here.”
That change in pronoun can pose problems. The system is built from what others can collect, such as messages, recorded habits, photographs, jokes, and remembered actions. It then represents a point of view, as if these fragments had recovered the inner life that once gave them meaning.
What appears to be preservation is, in reality, creation. Each response combines data about the deceased with selections from the family, input from the bereaved, predictions from a model, and design decisions from a company. The response sounds as if it came from a single person, but its authorship may be shared among many.
What is being kept alive?
Is it healthy to keep talking to the deceased ? Grief has never demanded silence. People talk at gravesites, write letters to the dead, repeat shared rituals, and imagine what a loved one would say. Maintaining a connection can be part of the grieving process.
The most difficult question is not whether people should maintain a relationship with the deceased. It is precisely what that act keeps “alive.”
Relationships help shape who we are. We become partners, parents, children, and friends through how others recognize and interact with us. Death takes a person from us, but it also disrupts a part of the survivor’s own identity. A parent who loses a child also loses the voice that made their parenting a vibrant exchange.
Inability to ‘turn the page’
A thanabot can recover some of that recognition. It can remember an anniversary, repeat a family joke , or call someone “Mom” again . But the service isn’t limited to recreating a version of the deceased: it can also preserve a version of the living person who existed within that relationship. This, of course, can offer comfort at first.
However, it’s possible that this keeps the grieving person anchored to a previous version of themselves, organized around a relationship that death has transformed. Companies could be selling access to the deceased and, in doing so, access to a previous version of the customer.
Who decides who the deceased become?
A thanabot can say things the deceased person never said. This might seem appealing, but ethical difficulties can arise precisely there. In 2025, the father of Joaquin Oliver , a 17-year-old killed in the 2018 Parkland school shooting, created an AI-powered avatar of his son and allowed a journalist to interview it. The avatar answered new questions using Joaquin’s voice and publicly defended his anti-gun activism.
The same question arises in the private sphere, within families. Who has the right to authorize a reconstruction: the spouse, a parent, all the siblings, or only the person themselves before their death? Love may motivate the reconstruction, but it does not grant unlimited authority over another person’s identity. And consent is only the beginning.
A thanabot can change over time. A software update can alter its vocabulary, its sense of humor, its “memories,” or its political leanings. Family members can upload new material, allowing it to evolve beyond the life it was trained to live. At some point, preservation gives way to interpretation.
From interpretation to invention
Interpretation then becomes invention. Ownership is therefore divided from the outset. The family can contribute the archive. The company controls the model, the interface, and the servers. The deceased person contributes the identity that gives the product its emotional and commercial value, but can no longer renegotiate the agreement. Companies can decide which version appears, what it remembers, and how long it remains available. Higher-quality memories may be reserved for a premium level . A familiar voice may change after an update. The clone may disappear when the company closes.
The psychoanalyst Jacques Lacan distinguished between biological death and symbolic death, which occurs when a person disappears from collective memory. Resurrection by subscription, as we suggested in another study , introduces a third possibility: digital death occurs when payments cease, the provider closes, or the model is withdrawn.
The power of the company behind a beloved image
In this case, ethics are not abstract. They are integrated into the service itself: what consent counts, who is responsible for distortions, how grief is assessed, and whether a family can pause, export, or permanently delete a reenactment without being forced to rent it forever.
The dead have always spoken through their belongings, rituals, and the people who remember them. What’s new is the animated, pre-generated response, along with a company’s power to edit, interrupt, and profit from it. As soon as mourning becomes dependent on server uptime, someone else gets to own the next line.
Author Bios: Vitor Lima is Associate Professor of Marketing at ESCP Business School and Russell Belk is Professor at York University, Canada
