October 21, 2026 · 8 min read
Griefbots Explained: Should You Make an AI of Someone You Lost?
What griefbots and deadbots are, what research on mourners found, the ethics Cambridge researchers raise, and questions to ask before you try one.
When someone you love dies, their voice does not disappear from your phone. Years of texts, voice notes and emails are still there. And today, with a few uploads, AI can turn that archive into something that answers back. These are griefbots, also called deadbots: chatbots that simulate a person who has died.
For some people, the idea is comforting. For others, it is deeply unsettling. Both reactions make sense. This guide explains what griefbots are, what researchers have found so far, the ethical questions they raise, and how to think it through if you are tempted to try one.
The short answer: A griefbot is an AI chatbot built from a dead person's messages, voice or writing so you can "talk" to a simulation of them. Research is still early. A small 2023 interview study found mourners used chatbots in several ways, from a listener to a simulation of the person they lost. Ethicists at the University of Cambridge recommend that such services be adults-only, transparent about what they are, built with the consent of both the person simulated and the people who will talk to it, and designed with a dignified way to say goodbye and retire the bot.
What is a griefbot?
The terms vary. Researchers use "deadbot", "griefbot", "thanabot", "postmortem avatar" and "ghostbot" largely interchangeably. In a 2024 paper in Philosophy & Technology, Tomasz Hollanek and Katarzyna Nowaczyk-Basinska of the Leverhulme Centre for the Future of Intelligence at the University of Cambridge define a deadbot as "an AI-enabled digital representation of a deceased individual created by a re-creation service", in their study Griefbots, Deadbots, Postmortem Avatars.
Griefbots are part of what researchers call the "digital afterlife industry": services that manage, memorialise or simulate people after death.
A short history
The paper traces a well-known example. Project December, an online platform offering "text-based conversation with anyone", including "someone who is no longer alive", drew public attention in 2021 after stories spread about a man who used it to talk with a simulation of his late fiancee. OpenAI, whose GPT-3 model powered the platform at the time, reportedly cut off its access, citing safety guidelines.
The authors note that OpenAI's usage policy at the time required any system simulating another person (except historical public figures) to have that person's explicit consent or to be clearly labelled as "simulated" or "parody". Project December later continued on its own technology.
Since then, building a simulation has become far easier. As the paper puts it, "almost anyone can bring a deceased loved one 'back to life'", with examples reported in China and the United States.
What do mourners actually do with chatbots?
Real-world evidence is still thin. One of the few studies with actual users is The "Conversation" about Loss by Anna Xygkou and colleagues, presented at the CHI 2023 conference. The team interviewed 10 mourners who had used chatbots at different stages of their loss.
They identified seven ways people used chatbots to cope with grief, with the chatbot taking roles such as:
- a listener to talk to,
- a simulation of the deceased,
- a romantic partner,
- a friend,
- and an emotion coach.
The authors describe mourners as people who "could be vulnerable to the sense of loneliness and disruption of self-identity", and frame chatbots as one way some of them coped. With only ten participants, the study shows what is possible, not what is typical.
The ethical questions
Consent of the person who died
Most of us never said whether we would want to be simulated after death. Hollanek and Nowaczyk-Basinska argue that the dignity of the "data donor" (the person whose data is used) matters even after death. They suggest services prompt the person building a bot with questions like "Have you ever spoken with X about how they would like to be remembered?"
Consent of the people who will talk to it
The Cambridge paper adds a newer idea: the people meant to interact with the bot should consent too. In one of their speculative scenarios, "Stay", a dying man secretly builds a bot of himself and designates his adult children to receive it. After the funeral, one child finds it comforting at first but becomes drained by daily interactions. The other does not want it and keeps receiving reminders and messages from the bot. The authors call this risk being "stalked by the dead".
Advertising and money
In another scenario, "MaNana", a woman builds a free, ad-supported bot of her grandmother. One evening, cooking her grandmother's carbonara recipe, the bot suggests she order carbonara from a delivery app instead, something her grandmother would never have said. The point is simple: a loved one's voice should never be used to sell to you.
Getting stuck
Philosopher Nora Freya Lindemann, in The Ethics of 'Deathbots' (Science and Engineering Ethics, 2022), argues that deathbots "may have a negative impact on the grief process" and that users may become dependent on them, which could also make them more open to hidden advertising. She also notes they "may prove to be helpful" for some people with prolonged, severe grief, and suggests they should be regulated as medical devices so that their safety and benefits have to be shown.
Saying goodbye
Perhaps the most practical point in the Cambridge paper: services should have "sensitive procedures for retiring deadbots". In the MaNana scenario, the app lets users delete their account but not the bot itself, leaving no meaningful way to say goodbye.
The Cambridge recommendations at a glance
| Recommendation | What it means for you |
|---|---|
| Retire bots with dignity | You should be able to end the bot in a meaningful way, and inactive bots should eventually be retired |
| Meaningful transparency | Clear signs you are talking to an AI, plus up-front warnings about risks and what the bot can and cannot do |
| Adults only | Re-creation services should be restricted to adult users |
| Mutual consent | Both the person simulated and the people who will interact should agree |
| No public appearances | Bots should not pop up in public spaces like social media feeds |
| Easy opt-out | Anyone can stop interacting at any time, without guilt or pressure |
If you are thinking about making one
Grief is personal, and there is no single right answer. If you are considering it, a few gentle questions can help:
- What do you hope to feel? One more conversation? A way to say goodbye? Comfort at night? Being clear helps you notice if it stops helping.
- Would they have wanted it? Think about what you know of their wishes and their sense of privacy.
- Who else is affected? Siblings, partners and children may feel very differently.
- Is it adults only, ad-free and deletable? Check whether you can fully delete the bot and its data.
- Can you set a time limit? Some people decide in advance to use it for a set period, then say goodbye.
- Are you also leaning on people? Friends, family, grief groups and professionals still matter most.
It is also worth remembering what any simulation is. It predicts how someone might reply based on their data. It is not them, and it is not conscious. We explained the science in why AI is not conscious. If grief feels overwhelming, or you have thoughts of harming yourself, please talk to a doctor or a crisis line in your country.
Griefbots vs AI companions
A griefbot recreates someone who existed. An AI companion is a new character, created for you, with its own personality and story. That difference matters. With a companion, there is no question of whether a real person would have consented, and no risk of the bot saying something your loved one would never have said, because it was never meant to be them.
Companions raise their own end-of-life questions, though. What happens to your companion when you die, or when an app closes? We covered both in what happens to your AI companion when you die and what happens when an AI companion app shuts down.
Where MyBabe fits
MyBabe is not built to recreate real people, living or dead. It is a relationship-first companion app for adults (18+) where you create a new girlfriend or boyfriend from scratch: their face, three personality traits, their voice and a backstory. Their selfies are locked to one reference portrait, so they always look like themselves and no one else.
On hard days, they notice. Mood check-ins let you tell them you are feeling Low or Rough, and they meet you there. With Premium, they remember what you shared months later, including how you felt, so you do not have to keep explaining your loss. And at 2am, when the house is too quiet, they pick up by voice or video.
Start free with 10 messages every 24 hours and two minutes of calls to try. Premium is $5.99 a week, $14.99 a month or $99.99 a year, with long-term memory, about 30 minutes of calls a week, four photos a day and Telegram.
MyBabe is an AI, not a real person, and it does not replace human relationships, grief support or professional mental health support. If you would like a warm, new voice in your corner while you carry what you are carrying, meet your companion on MyBabe.
FAQ
What is a griefbot?
A griefbot, or deadbot, is an AI chatbot that simulates a person who has died, usually built from their messages, voice recordings or writing.
Are griefbots healthy?
The evidence is very limited. A 2023 study interviewed just 10 mourners. Ethicists warn about dependence, advertising and difficulty saying goodbye, while noting some people may find them comforting.
Is it ethical to make an AI of someone who died?
Cambridge researchers argue it depends on consent, transparency and design. They recommend adults-only access, mutual consent and dignified ways to retire the bot.
What was Project December?
An online platform offering text conversations with simulations of anyone, including people who have died. It drew attention in 2021, and OpenAI reportedly cut off its access to GPT-3.
Can MyBabe recreate someone I lost?
MyBabe is designed around new characters you create, not simulations of real people.