quiddity
Griefbots and Authentic Voice Preservation: What Families Should Know
What griefbots are, what the research says about their psychological effects, and how authentic voice preservation differs — ethically and practically.
The grief impulse is ancient and entirely human: we want, against all reason, one more conversation. One more story. The sound of a voice saying something new.
Griefbots — AI systems trained on text messages, emails, and recordings to simulate ongoing conversation with someone who has died — offer a version of that. The technology is real, accessible, and growing. So are the questions about it.
This is not a piece against grief or against technology. It is an attempt to give families the clearest possible picture of what griefbots actually are, what research suggests about their effects, and where the ethical and practical lines fall between simulating a person and preserving what they actually said.
This article is for informational purposes only and is not a substitute for grief counseling or professional mental health support. If you or someone you love is struggling with loss, please reach out to a qualified mental health professional.
What is a griefbot?
A griefbot is an AI system trained on a deceased person’s digital footprint — their text messages, emails, social media posts, voice recordings — so that living people can hold new conversations with a simulation of them. Also called “deadbots,” “digital ghosts,” or “AI afterlife” services, these systems generate responses the deceased never actually gave, based on patterns learned from what they left behind.
Several platforms now offer versions of this. Some were built specifically for grief. Others — such as Replika, originally marketed as a general AI companion — found this use case arrive unbidden, as bereaved users began directing them toward lost loved ones. The field is growing quickly enough that researchers at the University of Cambridge’s Leverhulme Centre for the Future of Intelligence called for regulatory safeguards in 2024, noting that existing design practices left users vulnerable to psychological harm and “unwanted digital hauntings” (Cambridge, May 2024). The ethical debate has moved from academic journals into mainstream media.
What does the research say about the psychological effects?
The honest answer is that rigorous research is still limited, and the findings so far are genuinely mixed.
Early exploratory research is beginning to appear. A pilot involving 10 participants who used AI simulations of deceased loved ones — covered by Scientific American in November 2025 — found that participants reported high levels of comfort from the simulations, and some showed increased rather than decreased social engagement with others. The sample is far too small to draw broad conclusions, and this is best treated as a preliminary signal rather than evidence of general effects.
The same body of research also identifies real risks. Psychologist Mary-Frances O’Connor warns that highly lifelike simulations of the deceased may interfere with the neurological relearning grief requires: the gradual process of accepting, at a neurological level, that someone is permanently gone. And research suggests that between 7 and 10 percent of bereaved adults may develop prolonged grief disorder — a condition in which grief does not ease with time (Newsweek, 2026). Whether ongoing AI simulation of the deceased increases that risk for vulnerable individuals remains an open question, not yet clearly established by the available research.
What is the posthumous consent problem?
One of the sharpest questions griefbots raise is whether the person who died ever agreed to be simulated.
In many cases, the answer is unclear or no. A family member may create a simulation from the deceased’s messages and recordings — data the deceased generated for entirely different purposes — without the person ever having been asked whether they wanted to be reconstructed this way. Researchers at the University of Cambridge’s Leverhulme Centre for the Future of Intelligence flagged this as a central problem in their 2024 call for safeguards: they proposed that services should prompt users to consider whether the deceased had expressed preferences about how they wanted to be remembered, and build in mechanisms for “dignified retirement” — ways to bring an ongoing simulation to a meaningful close rather than leaving it running indefinitely (Cambridge, May 2024).
The researchers also noted that when the deceased had signed a contract with a service before death, surviving family members may have no ability to suspend the simulation — even if interactions had become distressing.
How does authentic voice preservation differ from griefbot simulation?
The clearest distinction is timing and agency.
Griefbots are created after a person dies, from data they left behind, to generate responses they never actually gave. Authentic voice preservation is something different: recording a person’s actual words, in their actual voice, while they are alive and willing to share them.
| Griefbot (post-death AI simulation) | Voice archive (recorded while living) | |
|---|---|---|
| When created | After the person’s death | While the person is alive |
| Whose decision | Often created by survivors, without prior consent from the deceased | The person decides what to share and when |
| What it produces | New AI-generated responses, based on patterns from the person’s past data | A fixed record of what the person actually said |
| How it works | An ongoing, open-ended simulated conversation | A permanent archive, revisited but not extended |
| What it preserves | A generative model of how the person might respond | The person’s actual words, cadence, and voice |
Recording while someone is alive has its own limits — not every family will have started in time, and even a willing parent may share only part of their story. The table above describes the distinction in approach, not a guarantee of outcome.
A voice recording of a parent telling the story of how they met, or what they were most afraid of at twenty, or what they want their grandchildren to know — is not the same as an AI trained to guess what they might say next. One is a record. The other is a model.
What should families consider before deciding?
Three questions are worth sitting with before turning to any grief technology. The answers won’t be the same for every family — but the questions themselves cut through the noise.
Did the person express how they wanted to be remembered? This is the consent question, and it matters most. If someone said, before they died, that they would not want to be simulated — or if they never had the chance to say anything either way — that context belongs in any decision about creating a simulation of them.
Who is this for, and what are they hoping it will provide? A tool that helps someone process a loss may serve very differently than one that allows them to avoid processing it at all. These are not the same outcome, even if the interface looks the same.
What can still be captured while there is time? This is the question that tends to get asked too late. Recording a parent’s actual voice, stories, and memories requires them to be alive. That window is usually longer than families expect — and shorter than they hope.
How to record a parent’s voice while they are still here
Recording a parent’s voice — their actual stories, in their actual words — is something anyone can start today, with nothing more than a phone and a single question. Here is what tends to make it go well.
Ask first, and make clear there is no pressure. Record only when the person is willing and has said so. Some stories may be ones they are not ready to tell; that is entirely their right.
Watch for tiredness or distress and stop — or simply pause — when they signal it. A short session that ends on their terms is still a good session. The goal is a genuine conversation, not the extraction of content. The most valuable recordings tend to come from moments where the parent forgets the phone is there, and it feels less like an interview and more like an ordinary afternoon.
How Quiddity approaches memory preservation
For families who want a guided way to capture a parent’s voice and stories while they are still here, Quiddity offers that. It records the actual person — their own words, in their own voice, unedited — through a guided conversation, and keeps the recording permanently for the family members they choose to share it with. There is no AI reconstruction. There is only what was actually said, kept exactly as it was.
If this is something your family is thinking about, it is worth exploring while the window is still open.