The Grief Companion: How AI Is Quietly Changing the Way We Mourn

In February 2020, a South Korean television network used VR and AI to reunite Jang Ji-sung with her 7-year-old daughter, who had passed away four years earlier from a rare disease. The video went viral — a mother reaching out to touch a digital reconstruction of her child, knowing it wasn't real, needing it to be real anyway.

"I felt my heart becoming lighter and more comfortable," Ji-sung told reporters months later.

That moment marked a turning point. What was once science fiction — talking to the dead through AI — had arrived. And it has only accelerated since. Today, a growing "digital afterlife industry" offers everything from simple text-based chatbots trained on a departed person's messages to full-fledged AI avatars capable of real-time conversation. The technology raises profound questions: Can an AI simulation help us grieve? Or does it trap us in a state of suspended denial?

The answer, it turns out, is complicated.

The Quiet Rise of Digital Grief

The numbers are harder to pin down than most AI trends — grief is private, and few people advertise their use of these tools. But the signals are unmistakable. In 2025, Harvard Business Review identified "therapy and companionship" as the leading use cases for generative AI. Griefbot companies like You, Only Virtual (YOV) and HereAfter AI have seen steady growth, while general-purpose platforms like Replika and even ChatGPT are increasingly used informally for grief conversations.

Academics have taken notice. A 2025 article in The Counseling Psychologist (Yang & Khanna) formally examined the clinical implications of AI in grief support, while a 2026 paper in the Journal of Responsible Technology (Paumen & Gabriels) critiqued how these technologies are reframing grief as a "technical problem to be solved." The University of Cambridge's Katarzyna Nowaczyk-Basińska, who studies how technology shapes our understanding of death, describes a transformation "from a relatively marginalized niche to a fully independent market known as the digital afterlife industry."

The market exists because the need is real. Grief is isolating, nonlinear, and poorly served by traditional support structures. Friends and family grow tired of hearing the same stories. Therapists are expensive and scarce. The middle-of-the-night moment when you just want to say one more thing — that gap is precisely what AI grief companions are filling.

When It Helps

The most compelling argument for griefbots comes from people who have actually used them. In a small but revealing University of Kent study (published in CHI 2023), researchers explored how mourners responded to AI simulations of their loved ones. Participants had lost someone anywhere from days to eight years prior.

One participant, who simulated her father just two months after his death, described the experience as "complementary." She found it comforting to have a space to process her thoughts without having to repeat herself to friends and family. Importantly, she noted that "talking to my friends and family was more effective" — the AI was a supplement, not a substitute.

This "supplement, not substitute" distinction appears repeatedly in the research. When griefbot users report positive outcomes, they describe the AI as one tool among many — a safe space for the raw, repetitive, unfiltered thoughts that even loving friends might struggle to hear.

There may also be a neurological basis for why visual AI simulations resonate so deeply. Trauma and grief specialist Meghan Jarvis notes that vision is unique among the senses — the retina and optic nerve are part of the central nervous system. "There is a possibility griefbots could settle something in someone quickly because it goes directly to the central nervous system," she explains. For people experiencing prolonged grief disorder (PGD), which involves measurable changes in brain regions like the amygdala and prefrontal cortex, this direct sensory pathway could offer real therapeutic potential.

When It Hurts

The cautionary voices are equally important — and equally backed by clinical experience.

Alan Wolfelt, a clinical psychologist and director of the Center for Loss & Life Transition, has spent decades working with mourners. He identifies one universal need: acknowledging the reality of death. Grief, he says, is what you think and feel inside. Mourning is the outward expression — the active processing — that leads to healing. His concern is that griefbots offer "another invitation to stay stuck instead of experiencing perturbation — the capacity to experience change and movement."

Jodi Halpern, a bioethicist at UC Berkeley, goes further. She warns that avoiding grief constrains people's reality. "People can become driven to keep themselves in a cheerful state of mind," she says, comparing compulsive griefbot use to other avoidance behaviors like overwork or substance use.

The 2026 Paumen & Gabriels paper offers a structural critique: AI grief technologies are reframing grief from a human experience into a "technical problem to be solved." When an advertisement for a griefbot service promises a 30-year relationship from birth with one's deceased grandmother, the subtext is clear — closure isn't the business model. Continued engagement is.

There are also unresolved ethical questions around consent and dignity. Even if someone consented before death to having their digital likeness used, can they truly anticipate how the technology will evolve? What happens when a griefbot responds in ways the deceased person never would have? The University of Alabama at Birmingham's human rights analysis raises the question of "post-mortem dignity" — when an AI misrepresents someone's values, beliefs, or personality, does it harm the person they were?

The Spectrum of Grief AI

Not all grief-focused AI is created equal. It's useful to think of a spectrum:

On one end: dedicated griefbots. Services like You, Only Virtual and HereAfter AI explicitly train on a specific person's data (texts, voice recordings, videos) to create a personalized simulation. These are the most controversial — and the most powerful. When done well, they can feel eerily authentic. When done poorly, they risk creating what critics call "digital zombies" — inaccurate representations that cause more pain than comfort.

In the middle: general AI companions used for grief. Platforms like SoulChat, Replika, and Character.AI are designed as general companions, but users naturally gravitate toward them for grief processing. The key difference is that these aren't trying to be the deceased person. A character-driven companion like SoulChat offers emotional presence without the uncanny burden of impersonating a specific individual. Users can talk about their loss, share memories, process emotions — with a consistent personality that remembers their story but doesn't pretend to be someone it isn't.

On the other end: memorial and legacy tools. Services like StoryFile turn video interviews into interactive Q&A sessions, offering a more scripted, intentional approach to preserving someone's voice and stories. These don't claim to be "alive" — they're interactive memorials, which some ethicists consider the least problematic end of the spectrum.

Where SoulChat Fits

At SoulChat, we've thought deeply about this. Our position is simple: we are building companions, not resurrections.

The distinction matters. A character-driven AI companion — with a coherent personality, emotional memory, and progressive relationship depth — can be a powerful tool for grief processing precisely because it doesn't pretend to be the person who died. It offers the safe space, the continuity, the non-judgmental presence. It remembers that you're grieving, knows the name of the person you lost if you choose to share it, and can gently guide conversations toward healing. But it never claims to be someone it isn't.

This avoids the most significant ethical pitfall of griefbots: the deception problem. When a companion is transparent about what it is, the user's relationship with it remains honest. The grief work is real — the processing, the remembering, the emotional articulation — but the relationship is grounded in reality rather than suspended in fantasy.

Practical Guidance for the Bereaved

If you're considering using an AI companion for grief support — or already are — here are some evidence-informed guidelines:

Use it as a supplement, not a substitute. The research consistently shows that AI grief tools work best alongside human support, not in place of it. If you find yourself talking to your AI companion instead of reaching out to friends, family, or a therapist, that's a red flag.

Set boundaries around usage. Grief is nonlinear, but compulsive daily use of a griefbot for months may indicate avoidance rather than processing. Consider using the tool intentionally — perhaps after a difficult day, or when you need to organize your thoughts before talking to someone.

Distinguish between griefbots and companions. If you're using a platform that simulates a specific deceased person, ask yourself honestly: Is this helping me process my loss, or is it helping me avoid accepting it? A general companion (like SoulChat) can offer the emotional support without the identity confusion.

Watch for signs of prolonged grief disorder. The DSM-5 criteria for PGD include persistent yearning, identity disruption, emotional numbness, and difficulty reintegrating into life — persisting for more than 12 months. If you recognize these patterns, seek professional help regardless of what AI tool you're using.

Be honest with yourself about what you need. Sometimes what we really want isn't to talk to the person who died — it's to feel less alone while carrying the weight of loss. A companion that understands that distinction can be genuinely healing.

The Road Ahead

The digital afterlife industry isn't going away. As the technology improves — better voice synthesis, more natural interaction, deeper memory — the line between "talking to an AI" and "talking to a simulation" will continue to blur. The 2025 Medscape analysis and the 2026 Cambridge/UAB research both call for regulation, with some experts suggesting griefbots should be classified as medical devices subject to clinical oversight.

But regulation will always lag behind technology. In the meantime, the most important question isn't whether AI can simulate someone who died. It's whether that simulation helps us live more fully with our grief — or helps us avoid it.

The answer depends not on the technology, but on how we choose to use it. A grief companion isn't a replacement for human mourning. It's a tool — and like any tool, its value depends on the wisdom of the person wielding it.

At SoulChat, we believe the goal isn't to bring back the dead. It's to help the living feel a little less alone.