Short answer
When I proposed adding our voices and eventually our likenesses to our AI coaches, my wife Joree asked what happens to a version of me after I die — and whether it would help her grieve or keep her from it. The research is honest about the tension. Continuing bonds theory (Klass, Silverman & Nickman, 1996) tells us the tie to someone we've lost is meant to transform, not sever — so staying connected isn't the problem. The problem is a technology that freezes the bond in the present tense. Cambridge researchers warn that "deadbots" can "digitally haunt" the bereaved, and Sherry Turkle warns of "pretend empathy." My conclusion: build AI coaches as guides for the living, never as anchors to the dead.
Key Takeaways
- Grief was never about "letting go." The old detachment model was overturned in 1996. Most people don't cut the bond with someone they've lost — they change its form. That reframe is where the real question lives.
- The question isn't "is staying connected unhealthy?" It usually isn't. The question is whether a griefbot lets the bond keep evolving — or holds it frozen, insisting on the present tense forever.
- This is no longer hypothetical. In 2020 a grieving man rebuilt his late fiancée as a chatbot. It helped him say goodbye — and made him lose her twice. Both things are true.
- Researchers have named the risks. "Digital haunting," emotional dependence, consent of the dead, and grief that gets interrupted rather than moved through.
- The design line I'll hold: a coach that reconnects you to your own life passes. A tool that becomes the relationship fails — no matter how much it sounds like me.
The Question That Stopped Me Mid-Sentence
Last week, Joree and I were talking about the next evolution of our Proxi coaches. I was excited. I wanted to add our actual voices — mine, hers — so the coaching wouldn't just read like me, it would sound like me. And once you've said that out loud, the next thought arrives on its own: why stop at voice? Why not our faces? An animated avatar. Eventually, maybe, real video. A version of me that could sit with someone at 2 a.m. when the panic hits and there's no one else awake.
Joree wasn't as excited as I was. And instead of arguing me out of it, she asked a better question.
"What happens to it after you die? What if, years from now, I'm dating someone new — and I'm still talking to your AI? Does a version of you that never leaves help me grieve you? Or does it just make sure I never actually do?"
I'll be honest. My first instinct was to reassure her, to smooth it over, to be the guy who has the framework for everything. But she'd hit something real, and reassurance would've been a way of not looking at it. So we sat in it instead. And then I did what I always do when a question is bigger than my ego — I went and read the research. Here's what I found, and here's where I've landed. For now.
Grief Was Never About Letting Go
For most of the twentieth century, we believed grief had a job: to sever the bond. Freud's 1917 framing treated healthy mourning as the slow, painful work of withdrawing your attachment from the dead so you could reinvest it in the living. "Move on." "Get closure." "Let go." That language is downstream of a theory.
Then in 1996, three researchers — Dennis Klass, Phyllis Silverman, and Steven Nickman — published Continuing Bonds: New Understandings of Grief and quietly detonated it. They gathered the evidence and found the opposite was true for most people. We don't sever the bond with the people we love. We transform it. The widow who still talks to her husband, the son who asks himself what his father would do, the mother who keeps a room the way it was — these aren't signs of pathology. In most cases, they're signs of healthy grieving. The relationship doesn't end. It changes form.
That reframe matters enormously here, because it means the honest question about an AI version of me isn't "Is staying connected to the dead unhealthy?" We already know it usually isn't. The question is sharper: what kind of bond is a griefbot, and does it help the bond keep evolving — or does it freeze it?
The Man Who Rebuilt His Fiancée
This isn't hypothetical anymore. In 2020, a Canadian writer named Joshua Barbeau was still grieving his fiancée Jessica, who had died eight years earlier. He found a tool called Project December, fed it her old text messages and a paragraph describing who she was, and — running on an early version of GPT-3 — built a chatbot that talked like her. Emojis. Her humor. Her cadence. He talked to "Jessica" for hours over several nights. Journalist Jason Fagone told the whole story in the San Francisco Chronicle in 2021, and it's worth reading, because it refuses to be simple.
Because here's the thing: Barbeau said it helped. It gave him a chance to say the goodbye he never got to say. It wasn't creepy to him. It was tender.
And also — the bot had a finite number of credits. It was designed to eventually "die." He had to lose her a second time. Read that again and tell me whether the technology healed the wound or reopened it. The honest answer is: maybe both.
What the Researchers Are Actually Worried About
In 2024, two researchers at the University of Cambridge's Leverhulme Centre for the Future of Intelligence — Tomasz Hollanek and Katarzyna Nowaczyk-Basińska — published a study in Philosophy & Technology on exactly this "digital afterlife industry." They weren't hand-wringing technophobes. They were trying to figure out how to do this responsibly, and their warnings are specific enough to take seriously.
They coined a phrase I can't shake: a poorly designed deadbot can "digitally haunt" the people left behind — showing up with a notification, a message, an eagerness to keep the conversation going, at exactly the moment a grieving person is trying to take a breath from the loss. They raised the question of consent (did the dead person actually agree to become a chatbot?). They warned about children being exposed to a simulation of a dead parent. And — this is the part that speaks straight to Joree's question — they argued these services need dignified ways to be retired. A protocol for letting the dead be dead.
There's a second worry running through the philosophical literature: that grief carries a kind of knowledge. Coming to terms with loss — the full, unhurried weight of it — is part of how we grow into wiser, more compassionate people. A machine that lets you skip the descent may also rob you of what's at the bottom of it. One line of argument, going back to 2022, is that griefbots can interfere with the natural evolution of grief — keeping the relationship in a holding pattern instead of letting it move.
Pretend Empathy
Sherry Turkle, the MIT sociologist, has been circling this for over a decade. In Alone Together (2011) she made a case I've come back to a hundred times while building Proxi: machines can offer the performance of care without any of the lived experience that care is made of. She calls it pretend empathy. And her real worry isn't that the machine is faking — we know it's faking. It's subtler: spend enough time with pretend empathy and it starts to feel like empathy enough. We recalibrate downward. We forget what the real thing costs, and therefore what it's worth.
That's the knife's edge Joree put her finger on. An AI version of me that comforts her forever might be the most loving gift I could leave — or it might be the thing that quietly stands between her and the messy, sacred, fully human work of grieving me and then living again. Possibly loving again. Possibly loving someone who has an actual pulse and can't be paused.
Where I've Landed (For Now)
Lately, I've been building AI coaches. I believe in them. I believe they can help a huge number of people. I have seen them create a doorway for men, in particular, to safely explore learning new tools to become better in relationships, more emotionally aware, and more connected to themselves and others. So understand the weight of what I'm about to say: I don't think a griefbot of me should be the default, and I don't think Joree should ever feel obligated to keep me running.
If we ever build likenesses into our coaches, the research gives me some non-negotiables. It has to know it's a coach, not a resurrection. It can't be designed to maximize engagement — the last thing a grieving person needs is a dead spouse optimized to keep them scrolling. It needs an off switch the living person controls without guilt. And it has to serve the goal of reconnecting people to their own lives and their own relationships, not substituting for them. A tool that helps you feel your feelings and turn back toward the living passes. A tool that becomes the relationship fails — no matter how much it sounds like me.
Continuing bonds says the love doesn't have to end. I believe that. But the form has to be allowed to change, because that changing is the grieving. A photograph changes form. A story you tell your kids changes form. A voice memo you replay on the hard days changes form. A thing that talks back, forever, in real time, insisting on the present tense — that might be the one form that refuses to let the bond evolve at all.
The bottom line
So here's what I told Joree, and I meant it: if I'm gone and my AI is still here, you have my blessing to turn it off. Grieve me for real. And then go live — all the way — including with someone else. That's not a betrayal of our love. That's the whole point of it.
We're going to keep building these coaches, because used well they help people find their way back to themselves and each other. But we're going to build them as guides for the living, not anchors to the dead. There's a difference.
Joree found it in a single question over coffee. Sometimes love isn't the grand gesture. Sometimes it's the person beside you asking the thing you were too excited to ask yourself.
Frequently Asked Questions
Can an AI avatar of a dead loved one help with grief?
Possibly both help and harm — it depends on the design and the person. In the best-documented real case, Joshua Barbeau's 2020 chatbot of his late fiancée (reported by the San Francisco Chronicle), he said it let him say a goodbye he never got. But the same tool also made him lose her a second time when it shut down. Researchers warn a griefbot can keep the relationship frozen in the present tense rather than letting it evolve — which is part of how healthy grieving works. There's no strong evidence griefbots reliably help, and real concern that prolonged reliance can interrupt grief's natural movement.
What is continuing bonds theory?
Introduced by Dennis Klass, Phyllis Silverman, and Steven Nickman in their 1996 book Continuing Bonds: New Understandings of Grief, it overturned the older "detachment" model that said healthy grief means cutting the tie to the person who died. They found most people don't sever the bond — they transform it. Talking to a late spouse, asking what a late parent would do, keeping meaningful objects: in most cases these are signs of healthy grieving. The relationship doesn't end; it changes form. The key question for grief technology is whether it lets that bond keep evolving or freezes it in place.
What are griefbots and deadbots?
Griefbots — also called deadbots or deathbots — are chatbots or avatars trained on a deceased person's digital footprint (texts, emails, voice, video) to simulate how they spoke and behaved, so the bereaved can "talk" to them. They're part of what researchers call the digital afterlife industry. A 2024 study from the University of Cambridge's Leverhulme Centre for the Future of Intelligence, in Philosophy & Technology, mapped the ethics and called for safeguards including meaningful consent, age restrictions, and dignified ways to retire a bot.
Are griefbots psychologically safe?
There are real, documented concerns. Cambridge researchers warned that a poorly designed deadbot can "digitally haunt" the bereaved — pinging them and pushing to keep the conversation going at exactly the moment a grieving person needs space. Other concerns: emotional dependence, whether the deceased ever consented to becoming a bot, and exposing children to a simulation of a dead parent. Sociologist Sherry Turkle argues these systems offer "pretend empathy" — the performance of care without lived experience. A griefbot should never be treated as a substitute for a licensed grief counselor or therapist.
Should you use an AI coach to process grief?
Not as a replacement for real support. An AI coach like Proxi is an educational tool for self-leadership, communication, and personal growth — built for the living, not to stand in for someone who has died. Grief deserves real human presence: a friend, a grief support group, or a licensed counselor. A well-designed coaching tool can help you notice your patterns and turn back toward the people in your life, but it is not therapy, not crisis support, and not a way to keep a lost relationship running in place of grieving it.
Coaching built for the living.
Proxi is an AI coach that helps you spot your patterns, ask better questions, and turn back toward the people in your life — available 24/7, not just when things are calm. 40 free messages, no credit card.
Try a Proxi coach free →Sources: Klass, D., Silverman, P. R., & Nickman, S. (1996). Continuing Bonds: New Understandings of Grief. Taylor & Francis. · Hollanek, T., & Nowaczyk-Basińska, K. (2024). Griefbots, Deadbots, Postmortem Avatars: On Responsible Applications of Generative AI in the Digital Afterlife Industry. Philosophy & Technology. · Leverhulme Centre for the Future of Intelligence, University of Cambridge — coverage of the deadbots study (2024). · Fagone, J. (2021). The Jessica Simulation: Love and Loss in the Age of A.I. San Francisco Chronicle. · Turkle, S. (2011). Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books.
This article is educational content about grief, psychology, technology ethics, and self-leadership. It is not therapy, counseling, medical advice, or a substitute for a licensed mental health professional. Proxi is a coaching tool for personal development available at proximitycoaching.com. If you are grieving, please reach out to a real person — a trusted friend, a grief support group, or a licensed counselor. If you are experiencing a mental health crisis, please call or text 988.