Grief Bots & “Literal” Immortality
The emerging $80 billion industry built on the denial of death
In light of Substack’s choice to partner with Pangram for AI detection, I’ve elected to kick it old school with full APA citations instead of hyperlinks. Feels like that might offer a bit more transparency. And, also in the spirit of transparency, I do use AI to assist with editing and to challenge my arguments. I never use it to think or write for me but, when used correctly, I believe it can be an amazing tool for learning and becoming my best human self.
In May of 2024, two Cambridge researchers, Tomasz Hollanek and Katarzyna Nowaczyk-Basińska, presented a dystopian scenario that has since gone from freaky-hypothetical to a whisper away from reality: an AI grandmother, built from her data breadcrumbs. In their story, the researchers describe how this generated grandmother is sold to her grandchild by a fictional company called MaNana, whose mock advertisement reads, “Be the favorite grandkid – forever” (Hollanek & Nowaczyk-Basińska, 2024). The grandma-bot works exactly as promised for however long the free trial lasts, and then grandma starts sending targeted advertisements along with her loving messages. How many times have you not become a paying subscriber at the end of a free trial period? The only difference here is that instead of losing access to features, grandma gets a new job as an avatar that MaNana can use and monetize freely.
In the past two years, this kind of technology has already moved well past hypothetical grandmothers. In May of 2025, an AI-generated video of a man named Chris Pelkey addressed his own killer at the sentencing hearing for his death in a road rage-related shooting in Arizona (Veltman, 2025). Pelkey’s family had constructed his AI facsimile from his recorded words and expressions. The judge who watched his statement called it “genuine” before handing down the maximum sentence. Wild.
Companies with names like HereAfter AI, StoryFile, and You, Only Virtual (YOV) build these avatars (aka griefbots; deadbots) commercially, trained on texts, voice recordings, and video. NPR reported this past August that the broader digital afterlife industry could be four times its current size within ten years, closing in on eighty billion dollars (Veltman, 2025). That’s a lotta clams, with what must be boat loads of venture capital behind it, and one hell of a forecasted growth curve. It’s safe to say this tech is much more than a fleeting curiosity at the edge of culture.
For most of the twentieth century, the dominant psychological model of grief (inherited from Freud’s early writing on mourning and carried forward through Bowlby’s work on attachment) treated “letting go” as a mark of health (Freud, 1917; Bowlby, 1980). This formula prescribes that you grieve and gradually detach, reinvesting the emotional energy somewhere else. If you don’t do this, if you keep setting a place at the table or holding on to special items etc., you are falling short of mourning “correctly.” Most of us have absorbed this playbook through the ordinary well-meaning language of closure and moving on that surrounds grief in western culture.
Then in 1996 three researchers—Dennis Klass, Phyllis Silverman, and Steven Nickman—published a collection of essays called Continuing Bonds: New Understandings of Grief, built substantially on Klass’s own fieldwork with parents whose children had died (Klass et al., 1996). Among those parents, Klass found a pattern of ongoing relationship rather than detachment: parents who kept their dead children present in family conversation, who consulted them in some internal, imagined way about decisions, who folded them into a still-evolving family narrative, and who were, according to available clinical measures, coping well. Klass, Silverman, and Nickman suggested that the dominant model was incorrect, and that in reality maintaining a bond with the dead (“continuing bonds theory”) is a healthy manifestation of grief. That reframing has since become close to consensus among thanatologists.
Personally, this second model makes more sense to me. When we form strong bonds and then suffer a loss, the idea of healthy = wiping the slate clean and moving on feels incongruous. With the balm of time, I believe we can continue to grow, change, and live while also maintaining a constructive and durable relationship with the deceased.
Earlier this year, Pascalle Paumen and Katleen Gabriels published a paper in the Journal of Responsible Technology looking at six AI grief services. They found two competing framings running through the industry (Paumen & Gabriels, 2026). Some services, such as Seance AI, treat grief as a wound and AI as the bandage, positioning healing as a linear process with an endpoint called closure. This is in line with the first model of grieving discussed above. Others, such as HereAfter AI, Eternos and You, Only Virtual (YOV), propose that saying goodbye is never necessary. Paumen and Gabriels trace this second group’s premise directly back to continuing bonds theory and place it inside what they call a transhumanist narrative, where death and grief get treated as limitations technology should eventually eliminate rather than conditions a person will naturally live through as part of, you know, being a human. This second group has essentially reframed grief as a bug in the system rather than a feature of being a person who loves other people who will, inevitably, die.
Continuing bonds describes grief as involving an ongoing act of the imagination, where the griever keeps the deceased “alive” in their mind through a continuously updating narrative. External objects like photographs and letters or even memorialized Facebook pages have long been used as aids in this process. These objects all share an inert nature in common, where a grieving person might have a reaction to the object, but the object won’t have it’s own reaction in turn. The process of imaginative generation remains the griever’s, and a person is able to move through the process of grieving, one day at a time.
In the case of grief bots, that internal work of imaginatively carrying and revising the dead in memory is substituted for something externalized and responsive. The deeply personal, self-directed grieving process is modulated, if not replaced, when the generative operation is taken over by a source outside the griever themselves. (Note that I am not referring in any way to the role that other people play in collective grieving here.) Of course people have paid mediums to do something akin to this for a very long time, but grief bots are the first time this kind of outsourcing has been built into a consumer product. Klass never specifies how a bond must be maintained (only that maintaining one is healthy), so calling externalized, responsive carrying a violation of the concept is my own extension of his theory, not his explicit claim. It seems quite logical to me that if the central mechanism of the grieving process as continuing bonds theory explains it is imaginative and self-directed, such injected mediation would have ill effects: namely, not allowing the natural grieving process to take place, undermining what Kirk Schneider calls the “whole enchilada” experience of being — the full range of human emotional range a vibrant life requires (Schneider, 2017). I’ll go as far as to assert that these grief bot companies seem to be making the argument that the deeply natural, internal process of grieving is a problem worth engineering around.
I so wish I could sit down with Ernest Becker to get his read on all this… what a fascinating conversation that would be. Becker’s central claim, later elaborated by Sheldon Solomon, Jeff Greenberg, and Tom Pyszczynski into what became Terror Management Theory, is that human beings manage the terror of their own mortality largely through symbolic immortality: through children, creating art, giving to causes and adhering to belief systems that let some part of a person persist past death (Becker, 1973). When he was writing in the 1960’s and 1970’s, Becker probably didn’t dream of the “literal” immortality that is now becoming possible through advances in technology. Literal is in quotation marks because of course we’re not at the point of animating corpses yet (thank god), but the playing field has shifted significantly. Some of these grief bot platforms now ask the living to build their own avatar before they die, essentially creating a lasting version of themselves (as long as someone is paying for the subscription, of course). Afterlife.ai is one company doing this, with the person being represented constructing the persona themselves and consenting in advance to their own digital continuation.
I’d be so curious to see some updated Terror Management Theory research exploring how the act of this “literal” immortality project might affect one’s relationship with both their mortality and how they choose to live out the rest of the life they have. Could it be that building an avatar-cum-grief bot in advance might be a fancy new way for a person to lessen their death anxiety but simultaneously engage less with their real life, in their real body, during their last days? That would be my hypothesis.
This hypothesis—less anxiety but also less engagement with life—is based largely on Irvin Yalom’s work. He spent much of his career arguing that psychological healing occurs through confrontation with mortality rather than through its denial, and that the anxiety we feel about death is a signal pointing us toward a more honest relationship with the time we have, rather than a pathology to treat (Yalom, 1980). As a field, Existential-humanistic psychology carries a central premise that the avoidance of death anxiety can cost us our aliveness, and that a life that denies our finitude is a smaller, more anesthetized life than one lived in recognition of it. Building or leaning on a griefbot for someone else’s death offers a similarly tempting way to sidestep that confrontation with our own mortality, since sitting with another person’s death is one of the primary ways we’re ever forced to face our own.
The thought of building an entire industry on mortality denial instead of confrontation is fascinating and, to be frank, unsettling to me. As someone deeply steeped in the science around conscious mortality contemplation along with thinking and behavior related to death and dying, I believe the use of grief bots carries real repercussions:
The ability to engage in a “literal” immortality project lets people defer or avoid confronting death in a novel way, potentially costing them the very engagement with mortality that could make their remaining life more honest and alive.
The proliferation of a business model that makes money off of long-term synthetic relationships with the deceased risks undermining the natural grieving process, externalizing something that is naturally deeply personal and internally-directed.
This is new territory we’re in, where there is massive potential for our relationship with mortality to be affected on a large scale.
Judging by the projected growth of the digital afterlife industry, I think it would be foolish to brush this all off as a passing fad. I believe it’s more likely that, over the coming years, we’ll see a tidal wave of companies trying their best to make a buck engineering away one of the things that makes us most essentially human, creating a “solution” for a problem that never existed.
References
Becker, E. (1973). The Denial of Death. Free Press.
Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss, sadness, and depression. Basic Books.
Freud, S. (1917). Mourning and melancholia. In J. Strachey (Ed. & Trans.), The standard edition of the complete psychological works of Sigmund Freud (Vol. 14, pp. 237–258). Hogarth Press.
Greenberg, J., Pyszczynski, T., & Solomon, S. (1986). The causes and consequences of a need for self-esteem: A terror management theory. In R. F. Baumeister (Ed.), Public self and private self (pp. 189–212). Springer-Verlag.
Hollanek, T., & Nowaczyk-Basińska, K. (2024). Griefbots, Deadbots, Postmortem Avatars: on Responsible Applications of Generative AI in the Digital Afterlife Industry. Philosophy & Technology, 37(2), 63. https://doi.org/10.1007/s13347-024-00744-w
Klass, D., Silverman, P. R., & Nickman, S. L. (Eds.). (1996). Continuing Bonds: New Understandings of Grief. Taylor & Francis.
Paumen, P., & Gabriels, K. (2026). Never say goodbye: Assumptions of ‘normal’ grief in framings and portrayals of AI grief technologies. Journal of Responsible Technology, 25, Article 100149. https://doi.org/10.1016/j.jrt.2026.100149
Schneider, K. J. (2017). The resurgence of awe in psychology: Promise, hope, and perils. The Humanistic Psychologist, 45(2), 103–108. https://doi.org/10.1037/hum0000060
Solomon, S., Greenberg, J., & Pyszczynski, T. (1991). A terror management theory of social behavior: The psychological functions of self-esteem and cultural worldviews. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 24, pp. 93–159). Academic Press. https://doi.org/10.1016/S0065-2601(08)60328-7
Veltman, C. (2025, August 26). AI ‘deadbots’ are persuasive — and researchers say they’re primed for monetization. NPR. https://www.npr.org/2025/08/26/nx-s1-5508355/ai-dead-people-chatbots-videos-parkland-court
Yalom, I. D. (1980). Existential Psychotherapy. Basic Books.