A Conversation with the Dead

Robert Neimeyer was just shy of his 12th birthday when he heard his mom’s scream — the scream that ended childhood, he’d later call it.

Read more Who’s Writing the Curriculum Climate Activists Are Using to Influence Judges?

On a cold January morning, Neimeyer’s mother found his father’s lifeless body. He died from a self-administered dose of whiskey and barbiturates.

Neither Neimeyer nor any of his family members saw it coming. They knew his dad had been struggling to keep his pharmacy afloat and that he had been dealing with health problems. But the pressures weighed more heavily on him than they realized.

As the eldest of three, Neimeyer was expected to step into the shoes his dad left behind. Over the years, he would often dream of his dad. In these dreams, his dad was a figure in the attic of his home. Neimeyer would try to approach, and his dad would vanish. Once, his dad met his eyes and smiled, but that was all. He vanished, as usual, without a word.

Decades later, though, this changed.

Neimeyer spoke with his dad, or at least that’s how it seemed.

“My God, Bobby, there you are,” his dad began. “You have no idea how long I’ve yearned to reach for you, and yet I never knew how.”

It was what Neimeyer, by then the director of the Portland Institute for Loss and Transition, had been longing to hear. He found himself weeping by the end of the conversation.

But his dad’s arms couldn’t hold him in this tender moment. All Neimeyer had to show for the conversation was a screen with four or five pages of messaging. Yet he walked away feeling reconnected, like it was a conversation he could always come back to.

All it took was an AI chatbot — in this case, a “griefbot” — to bridge the gap left by death.

“Bob, of course I cannot be your father, but I can attempt to give voice to that father that perhaps you needed to hear from,” the griefbot had said. After sharing several paragraphs about his dad and explaining what he “would like to achieve” through the dialogue, Neimeyer began a conversation that he would later tell National Review was “very personal” and “surprisingly accurate.”

Sometimes called “deadbots,” these griefbots fall into the overarching category of “generative ghosts,” AI representations of a deceased family member or friend. Griefbots are specifically utilized to help people grieve, trained with everything from the deceased loved one’s social media accounts to biographical information inputted by the user.

An early prototype of a griefbot was generated in 2013, but the rise of ChatGPT in 2022 took the technology to a new level, further humanizing and personalizing AI conversations.

For some, griefbots might seem like nothing more than a disturbing plotline in a movie. A May 2024 episode of Star Trek: Discovery involves a character using an AI program designed off of his deceased grandmother to advise him and help him work through his pain.

Likewise, a 2013 episode of the British sci-fi television series Black Mirror — a graphic show in which new technologies unleash dystopian futures — offers a presentation of these so-called generative ghosts.

The Real Thing or a Harmful Substitute?

Griefbots are no longer confined to science fiction: For an increasing number of people, they offer what feels like a very real opportunity to grieve. But as AI technologies continue to develop, some psychologists worry that griefbots might provide a simulacrum of the grieving process that could harm, rather than help, the bereaved.

“If we continue to pretend our loved one is just absent and not permanently deceased, and that our relationship to the deathbot is a transformation of what was, and not the same relationship with who we loved, we are no longer living in reality, and no longer in a reality with our living loved ones,” Mary-Frances O’Connor, Professor of Psychology and Head of the Grief, Loss, and Social Stress Lab at the University of Arizona, told NR.

“While there might be limited uses of AI that helpfully give someone an opportunity to say goodbye, especially in the context of psychotherapy, deathbots have strong and concerning potential to offer the brain an impression that their loved one is still alive in the present,” she added.

O’Connor discusses the grieving process at length in her book The Grieving Brain. She specifically differentiates between the terms “grief” and “grieving.” Grief, on the one hand, is recurring: “You will experience pangs of grief over this specific person forever,” she said. On the other hand, grieving is a person’s relationship to the sensation of grief. It is a process with a trajectory, and it changes over time.

She stressed the human brain’s difficulty contending with grief and explained that the grieving process entails people reworking their “virtual maps” and adapting how they see the world. O’Connor is worried by the extent to which griefbots blur the line between past and present. Griefbot users must consider whether the bot is helping them recognize that their loved one is deceased or whether they find themselves believing that he or she is everlasting.

George A. Bonanno, Columbia University Professor of Clinical Psychology, likewise feared that griefbots could forestall a vital process for getting past a loss.

“It’s confusing our mind,” he told NR. “And our mind needs to know when a person is no longer living.” Despite the minimal research that has been done on griefbot use compared to other issues, Bonanno said he is “very, very concerned that it could be harmful.”

His trade book, The Other Side of Sadness, suggests that humans are far more resilient than they realize when it comes to grief.

“Most people cope with loss very well — I don’t think that’s fully understood,” he told NR. Bonanno said less than 10 percent of people who experience a loss suffer more seriously in the long run.

Due to the intimate nature of griefbots, he was also troubled by their potential to be used manipulatively by the tech giants running many of the top AI technologies.

“Whoever, whatever companies are making these things are obviously not doing it because they just like people and they want to help people,” he told NR. “They’re doing it to make money.”

Specifically, these executives could monetize people’s connection with their chatbot loved one once they have formed an attachment.

O’Connor also feared the potential for data harvesting from these heartfelt conversations. Once LLMs have a person’s patterns of thought, deep questions, fears, and so on, there is no way to take it back. Though griefbot companies claim that each user has his or her own closed database, there is still the potential for hacking or accidental leaks.

Neimeyer draws a different conclusion from the grieving process.

Rather than cutting off ties with the deceased loved one, he said it can be helpful to maintain them.

“We constitute our very sense of who we are, by whose we are,” he said. “And when we lose that person who provides a kind of anchor point for our life, who holds the mirror to our soul, . . . we have not only a rupture of the world in the form of the traumatic event of the loss, but also we have the rupture of the attachment bonds that give life much of its meaning.”

Read more Pressed by Trump, Thune Still Has Support Where It Counts

Saying the Unsaid

In a November 2025 article for Scientific American, author and independent researcher David Berreby told the story of using a griefbot to simulate a conversation with his deceased father.

“I said things to my father’s AI ghost that I never would have said to the real man, and I think doing so helped me clarify some of my version of our relations,” he said. “As I explored my take on our history, I felt my attachment to my version diminish. . . . I still thought I was pretty much right, but I found myself feeling more empathy than usual for my father.”

He ultimately compared the conversation to a game, video, or even an “imaginary chat” with his dad and said it had no negative effect on him, leaving him “optimistic about the dawning era of the digital ghost.”

Though Neimeyer also sees griefbots favorably, he believes they should be a companion for — not a replacement of — traditional therapy.

“I think that’s what we need is to train these AI agents as if we were supervising an eager and very, very bright young therapist who needed to balance caring and daring with their clients,” he said.

This is the mindset behind Working with Continuing Bonds in Grief Therapy, a book authored by Neimeyer and Edith Maria Steffen, Associate Professor in Counselling Psychology at the University of Plymouth, U.K. In it, he suggests that retaining a connection with the dead through AI could be helpful but requires more developed guidelines to be cemented as a professional practice.

“I think that what we need is a collaboration of human intelligence and carbon-based compassion with silicon-based intelligence and collaboration,” he told NR.

But even Neimeyer acknowledged that the agreeable nature of AI chatbots could become a “convincing refuge” for a grieving individual.

“We count on those we love the most and trust the most to give us a nudge in the direction of doing things that are hard but necessary in life,” he said. “And there are things that are hard but necessary in grief that go beyond being held and cared for and comforted.”

As a result, Neimeyer has spent years creating a specially tailored app for clinics and grief counselors called MeaningBridge. The app is presented as “a quiet presence between you, the people you love, and the clinicians who care for you.” Patients create accounts and have access to everything from guided journaling and voice entries to connection with a griefbot based on a deceased loved one.

Nevertheless, many individuals and organizations are far less eager to promote griefbot use.

An article published by the Hastings Center for Bioethics and authored by Craig Klugman, Vincent de Paul Professor of Bioethics and Health Humanities at DePaul University, discusses several of the more problematic implications of griefbots. Klugman said it is vital to consider the unique ways griefbots could affect different age and cultural demographics.

“Complicated grief is the more likely outcome for children who, depending on the developmental stage, might see death as an impermanent state,” Klugman said. “A child who can see a parent’s AI ghost might insist that the parent is alive.”

In a 2023 study, “The ‘Conversation’ about Loss: Understanding How Chatbot Technology was Used in Supporting People in Grief,” Neimeyer and six other researchers studied ten people — one woman and nine men — who were each using a chatbot to grieve the loss of a loved one.

The study highlighted the well-known fact that people tend to distance themselves from family and friends after a loss. The researchers found that “renewal of this bond” with the deceased individual can “support social connectedness” with him or her “in life-affirming ways.” Specifically, correspondence like letter writing and “imaginal written dialogues” were widely regarded as one of the best ways to work through grief.

Participants said the chatbot provided comfort that others weren’t able to give — or they were too insecure to ask for — and many found themselves leaning on it in moments of solitude.

“It was late at night, when I didn’t have anything else going on. I was bored, maybe sad, looking for answers; and so it was during those really quiet sort of times when I was alone, that I’d open chatbot,” one participant reflected. “And I think in those moments, for me personally, those were the most beneficial time [sic] to use it.”

But some participants admitted that the griefbots offered superficial or irrelevant advice rather than the guidance that would have been most beneficial in the long run.

“The main problem is she can’t remember anything,” one participant said. “You can talk to her, and two sentences later, she didn’t know what we are talking about.”

Posthumous Privacy

The issue of griefbots has also raised questions about digital legacy and end-of-life support for people suddenly left with their deceased loved one’s Google, Instagram, Spotify, and other accounts.

University of Birmingham Associate Professor in Law Edina Harbinja and several of her colleagues laid out a comprehensive framework for addressing this complex issue in “Model Rules on Succession and Access to Digital Remains.”

The rules were approved in July 2026 by the European Law Institute and centered specifically around handling someone’s digital assets after his or her death. Though they do not have legal power on their own, the rules are designed to inform and advise legislators.

They specifically advise that personal digital “assets” be passed down via succession, while “[p]ersonal digital remains, closely connected to the deceased’s identity, privacy, and dignity” fall under a “right of access.”

The issue is so complex that Harbinja devoted an entire book to it. Governing Digital Immortality: Artificial Intelligence, Deadbots and the Law comes out in September 2026 and raises deeper questions surrounding the issue of griefbots — questions like “Who has the right to create such simulations of the deceased?” or “What happens if someone is recreated without consent?”

Admittedly, the advancement of chatbot technology allows users without a background in software to easily customize and train chatbots, often without thinking twice about the consequences and deeper implications.

Regardless of how this technology is used, its popularity seems assured given that its adoption is driven by a deeply rooted human desire — the desire to convene with lost loved ones — that isn’t going away.

“I think the fact that people might be interested in the chatbots is simply because they’re now available,” Bonanno told NR. “Grief is just painful. There’s no way around it.”

Read more N.Y. Nuns Won’t Be Forced to Comply with Assisted-Suicide Law While Case Is Adjudicated