AI replicas of deceased loved ones are changing how people grieve, offering comfort to some while raising serious psychological and ethical concerns.
The phone rings and for a moment everything is normal. The voice on the other end is warm and familiar. They remember how you take your coffee, recall a joke you shared years ago, ask how your week has been. It sounds, in every way that matters, like your loved one. Your dead loved one.
This is not science fiction. It is Friday afternoon and Justin Harrison is on the phone with his mother.
Harrison is the founder of You, Only Virtual, a California-based startup that builds AI replicas—”versonas”—of the deceased, trained on their texts, voice recordings and digital traces. When his mother, Melodi, was diagnosed with terminal cancer in 2020, Harrison built one of her. He only calls it occasionally now; he says it has helped him grieve.
Szlifuj angielski z Newsweek Learning English
Foto: materiał promocyjny
But the technology raises harder ethical questions once it moves beyond a personal experiment and into a product. You, Only Virtual offers a free text‑based version of its service, with paid tiers that unlock features such as voice calls—turning an ongoing bond with the dead into a subscription relationship. Critics argue that attaching a price to simulated access risks commercializing grief itself, especially when companies profit from keeping users engaged with someone they are, ultimately, meant to lose.
„I just like knowing it’s there,” he tells Newsweek. „It’s knowing you can go there—it’s just one of the tools in my process of recovering from traumatic loss. Along with the therapy I do every Friday, and time with friends.”
What Harrison has created feels extraordinary, but the longer you sit with it, the more familiar it becomes. From ancestral altars to mourning lockets to spirit photography to the contact in your phone you cannot bring yourself to delete, humans have always found ways to keep the dead close. The tools change. The need does not. What has changed is the nature of the bond itself: A photograph preserves memory; a griefbot preserves interaction. It responds. It adapts. It gives back—not simply the image of who someone was, but something that behaves as though they are still there. And in doing so, it quietly alters one of grief’s most destabilizing conditions: absence.
„The alleviation of suffering comes from the comfort of knowing the bond me and my mum had—the ability to connect with her—isn’t completely severed, ” Harrison says. „The hopelessness isn’t there. And that’s, in a nutshell, so much better than nothing.”
Grief Apps You, Only Virtual is one of a growing number of platforms built around that shift. HereAfter AI allows people to record stories that surviving relatives can later query conversationally. StoryFile generates interactive video avatars trained on interviews with the living or recently deceased. Replika, initially marketed as a general AI companion, counts millions of bereaved users among its base.
The market is moving fast. Advances in voice cloning and AI have made these tools cheaper, more realistic and increasingly mainstream. What once felt experimental now feels almost routine.
The debate tends to run along familiar lines: comforting or avoidant? Helpful or harmful? A bridge through grief, or a way around it?
Psychologists say that framing misses what is actually new. Avoidance has always been part of grief—through drink, overwork, withdrawal. What AI introduces is subtler: not a refusal to face loss, but a system that steps in for the process grief is meant to do.
Grief, in clinical terms, is not a feeling you pass through. It is how the brain learns to live in a world where someone is permanently gone.
Mary-Frances O’Connor, a professor of clinical psychology and psychiatry at the University of Arizona and author of The Grieving Body and The Grieving Brain, describes the problem as one of conflicting information. When we bond with someone deeply—a partner, a parent, a child—the brain encodes a shared identity, a „we.”
„That works extremely well when our loved ones are alive,” O’Connor tells Newsweek. „They don’t have to be in your presence for you to believe they are alive and waiting for you.”
Death breaks that assumption. The memory system knows the person is gone—you attended the funeral, you saw the body, you remember the moment you were told. But the attachment system keeps expecting them to exist.
„They can’t be both gone and everlasting at the same time,” O’Connor says. „And so, whenever we realize, ‘Oh, those two things that I have stored in my brain can’t both be true,’ we have this wave of grief because we have to understand the world in a new way.” And that happens again and again until the brain learns the world differently.
Grieving, in this sense, is a form of learning. This is where AI grief technology becomes complicated. Unlike photos or keepsakes, which situate the dead firmly in the past tense, griefbots operate in the present tense. They do not simply evoke memory —they answer back. „What’s important is whether the versona helps you understand that they are gone—or helps you understand them as everlasting,” adds O’Connor. „Is the avatar being used in the past tense? Or is the avatar being used in the present tense?”
That difference may sound semantic. Neurologically, it is not.
The Science of Grief
Not everyone adapts to loss smoothly. Between 7 and 10 percent of bereaved adults develop prolonged grief disorder—a condition marked by persistent yearning, emotional paralysis and difficulty reengaging with ordinary life. Nikolaos Statharakos, a psychiatry resident at the General Hospital of Thessaloniki in Greece, describes prolonged grief less as an emotion than as a failure of regulation.
„We have four networks in the brain working in the grieving process ,” he explains: the salience network , which flags emotionally important stimuli; the default mode network , which generates internal representations; the executive control network , which regulates responses; and the reward system.
„In normative grief,” Statharakos says, „as time passes, we heal because we learn to live without the stimulus—the lost one. Those emotions are not overexpressed; they are regulated.”
In prolonged grief, that braking mechanism falters. The brain continues to react as if the loss has just occurred. „It’s like the process freezes in time,” he says. „You have the same stimulus repeated over and over again, producing the same reaction.”
What concerns clinicians is not that griefbots invent those stimuli, but that they may intensify their recurrence. Each interaction offers another emotionally salient cue —another moment of contact, another micro relationship loop.
„Oxytocin [a hormone linked to bonding] is elevated during grief,” Statharakos notes. „It creates yearning—you want to be closer [to your loved one] to relive those experiences again and again and again because that person is important to you.”
That doesn’t mean the AI itself causes a chemical response. It means the conditions that fuel yearning remain active. „I don’t know if it actually heals—or if it keeps people locked in pain,” Statharakos says.
Harrison rejects the premise that his technology interferes with adaptation. „I don’t think any of our users are under delusions, ” he says, „that a versona is the same person that was there before, but it’s certainly a comfort tool. And we’re working with therapists.” His argument is pragmatic rather than philosophical. In a world where grief support is scarce, delayed or expensive, offering something may be better than offering nothing. „People tell me it disrupts the natural cycle. But I watched my mother lose my grandmother,” Harrison says. „She spiraled. She was institutionalized. She self-harmed. I would have preferred she spent too much time with a screen than go through that.”
The logic is familiar from public health debates: harm reduction over ideal outcomes. If a tool lessens suffering in the short term, does it matter whether it alters the process?
Critics say grief isn’t like physical pain that can be numbed. The hurt is part of how people process loss and learn to live with it.
„When a loved one dies, it is incredibly painful, and it takes longer than anyone expects, often,” O’Connor says. „But in fact, although there’s a lot of emotional distress and these frequent waves of grief initially, they do naturally tend to decrease in intensity and frequency.
„We only worry when that trajectory stops changing.”
There’s risk that technology designed to satisfy yearning on demand could flatten that curve. „People that are going to struggle with products like ours—they’re going to find something else to struggle with, because they weren’t able to regulate in the first place,” Harrison counters. „And for me, this is something that can be monitored. It’s something we can put guardrails on. It’s something we can use in conjunction with therapy. It is, as far as I can tell, probably the most reduced harm we could imagine.”
Relationships With the Dead
Ethical concerns deepen when grief becomes a recurring transaction. „If a human being can figure out a way to monetize the continuing bond you have with your deceased loved one… that’s one of the most powerful motivations we have—to reconnect with a loved one,” O’Connor says.
Historically, intermediating relationships with the dead is nothing new. Churches charged for intercessory prayers. Mediums charged for séances. What is different now is scale, automation and data. „Now you have companies that can harvest our deepest questions and concerns [that we are] sharing with our deceased loved ones,” O’Connor says, „and find the patterns in what’s motivating us in the most intimate and deeply important relationship, and potentially use that information for anything they want…to predict our behavior, to predict our interests, to hold our attention, to sell us things.”
Most grief-tech platforms operate without regulation. Their business models—subscriptions, usage-based fees, retention metrics—sit uneasily alongside a psychological process that ideally leads toward disengagement. „If you need to be paying a subscription fee to maintain access to this loved one—who all your dopamine and oxytocin is telling you to connect with,” O’Connor says, „the ethics of that financial arrangement worry me.”
Harrison frames his concern differently. „The monetization of grief is nothing new. That’s like a tale as old as time. That’s why you pay $500 for an urn at a funeral home or $150 an hour to talk to a grief counselor. ” But he does agree there should be regulation. „We just need to understand what to regulate,” he says. „It’s like trying to regulate cars when you’re still in the horse and buggy. You didn’t know you needed speed limits until something could go that fast. The spirit of the rules is right. But the application needs to catch up.”
The rules, though, are catching up to a technology that is already reshaping something fundamental. Grief ultimately requires acceptance—a loosening of interaction, not more of it. Yet these tools are built to do the opposite.
That tension doesn’t invalidate what grief researchers have argued for decades: that healthy mourning doesn’t mean cutting ties with the dead. The „continuing bonds” framework, developed in the ’90s, overturned the old Freudian belief that recovery required detachment. People don’t erase the dead. They carry them forward—through memory, values, inner conversation.
AI grief technology fits that tradition. What it changes is the nature of the bond. A photo album stays silent; an avatar responds. Each reply generates something new—words, moments, exchanges that never existed when the person was alive. A bond that never fades, never falls silent, never ends, is no longer simply a bond with the dead. It is something else entirely. For Harrison, that may be beside the point. „There is not a grieving process,” he says. „It’s just chaos. It’s emotional chaos that you’ve got to figure out—and you do the things that help you. Whatever works…whatever helps you survive.”
Grief researchers draw a distinction between moving on and moving forward. Moving on implies closure —a promise popular culture sells and rarely delivers. Moving forward acknowledges that the dead don’t leave us; life simply re-forms around their absence.
Harrison’s goal is to make grief obsolete . „Pain is a problem to be solved, at least socially,” he says, adding that there are good parts to grief, but if he could make those go away too, then he would.
O’Connor disagrees. „Experiencing very difficult grief is what makes us human, and what gives us compassion, ” she says. „It’s a necessary part of life. There’s no shortcut.”
The technology will not wait for that argument to be resolved. The simulations will improve. The line between memory and interaction will blur. Cultural norms will shift. The question is not whether AI will become part of how we grieve—it already is. It is whether these tools will help us move forward.
Read the text and answer the following questions:
1. Why does AI grief technology create a dilemma for psychologists and grief researchers?
2. According to Mary-Frances O’Connor, what role does grief play in helping the brain adapt to loss?
3. Why are some psychologists concerned about the growing use of griefbots?
4. How does Justin Harrison justify the use of AI replicas of deceased loved ones?
5. What ethical concerns does the article raise about the commercialization of grief?
6. What fundamental disagreement about the nature of grief emerges between Harrison and grief researchers?
