Should You Make an AI Version of Someone Who Died? Questions to Ask First
From Everendium · Last reviewed October 7, 2026
Quick answer
There's no settled answer, and researchers are cautious. Cambridge ethicists who studied 'griefbots' called the area high risk, warning of unwanted 'hauntings,' hidden advertising in a loved one's voice, distress for children, and users who bond so strongly they become vulnerable to manipulation. Some scholars see possible comfort through 'continuing bonds'; others worry about dependence, denial, and prolonged grief. The effects on grieving people are still poorly studied. Before deciding, ask: Would they have wanted this? Whose messages and data would it use, and do you have the right to use them? Who controls it, and what happens if the company changes or closes? Are you in acute grief? Can you set limits and a plan to stop? If you're struggling, a human counselor is a better first step.
It’s now possible, with a few hours and some old messages, to build a chatbot that writes like someone who died, or to animate a photograph or clone a voice. For some grieving people the idea is comforting. For others it’s frightening. This guide doesn’t tell you what to do. It lays out what researchers say, what we don’t yet know, and questions worth asking before you decide. If you’re thinking about this for your own legacy rather than someone else’s, see an AI version of yourself after you die.
What are we talking about?
Researchers use terms such as griefbots, deadbots, and postmortem avatars for AI systems that imitate a person who has died, usually by learning from their texts, emails, voice recordings, videos, or writing. Cambridge researchers describe an emerging “digital afterlife industry” of companies offering such services, and note that rapid advances in generative AI mean nearly anyone with internet access and some basic know-how can now revive a deceased loved one.
What do researchers say?
The most-cited work comes from Cambridge’s Leverhulme Centre for the Future of Intelligence. Dr. Tomasz Hollanek and Dr. Katarzyna Nowaczyk-Basińska described the area as “high risk” and offered design scenarios to show what can go wrong. Their concerns include:
- Hidden advertising. A company could use a deadbot to promote products in the voice of a departed loved one
- Distress for children. A bot might insist a dead parent is still “with you”
- Vulnerability to manipulation. Because people may develop strong emotional bonds with such simulations, they may be particularly vulnerable
- Consent and rights. The researchers say designers should seek consent from data donors before they die, though they argue a ban based on non-consenting donors would be unfeasible, and that services should consider the rights of those who will interact with the simulation, not just the person recreated
- How it ends. They argue that methods and even rituals for retiring a deadbot in a dignified way should be considered
Later academic commentary describes both sides. A 2025 philosophy chapter notes the idea that griefbots could support the “continuing bonds” approach to grief, in which people maintain an ongoing sense of connection with the person who died, but also that vulnerable grievers face emotional hazards. The same literature includes proposals to regulate griefbots as medical devices under professional supervision, which others say would limit bereaved people’s freedom, and a suggestion that griefbots serve as a temporary form of commemoration instead of a long-term surrogate. A survey-based study of generations found caution across age groups, with concerns including emotional dependence, denial, and prolonged grief. A 2026 review notes the worry that users may come to prefer the bot over real relationships, which could undermine coping and adaptation to loss.
What don’t we know?
A lot. The Cambridge paper itself says that knowledge about the impact of AI-mediated grief on mental and physical health is still evolving. Most of what’s published is philosophical, ethical, or survey-based, not controlled clinical research, and we haven’t found studies showing whether these tools help or harm grieving people over time. Treat anyone who claims certainty either way with caution.
Questions to ask before you decide
About the person
- Would they have wanted this? Did they ever say anything about AI, recordings, or being “recreated”? If you don’t know, consider that you can’t ask
- Does it represent them fairly? An AI will invent things, and it may say what they never would have said
About the data 3. Whose words are you using? Texts and emails include other people’s messages. Texas law, for example, makes access to the content of a deceased person’s electronic communications depend on the user’s own direction, and terms of service can limit access, so even an executor can’t always read private messages; see can your family access your email? 4. Where does the data go? Find out whether a company keeps it, trains on it, or can delete it on request
About control 5. Who owns and controls the bot? What happens if the company raises prices, adds advertising, changes the bot’s personality, or shuts down? 6. Can you stop? Look for clear ways to delete the bot and your data
About you 7. Where are you in your grief? Early grief is a vulnerable time, and researchers emphasize that a griever’s vulnerability calls for careful risk assessment. If you’re struggling to function months later, see complicated grief 8. Are other people involved? Children, in particular, may be confused; see talking to children about death 9. Can you set limits? For example, a set number of sessions, a time limit, and a plan for when you’ll stop 10. Is there a human you can talk to? Tell a counselor, friend, or support group what you’re doing
Signs it’s time to stop or get help
- You’re avoiding people, work, or sleep to spend time with it
- You’re having trouble accepting that the person has died
- You feel worse after using it, not better
- You rely on it to make decisions
- You have thoughts of harming yourself. In the U.S., call or text 988 (Suicide & Crisis Lifeline) or call 911 in an emergency
What are gentler alternatives?
Many people find comfort in methods with fewer unknowns: re-reading old messages, listening to saved voicemails, watching home videos, writing letters to the person, making a memory book, or recording stories from family; AI can help organize these without pretending to be the person; see can AI preserve your life story?. A grief group or counselor can help you keep a bond without a simulation; see finding grief support.
What about the law?
Using a dead person’s voice or likeness raises questions of consent, privacy, and, in commercial settings, state law; see can anyone use a dead person’s voice or likeness?.
This is general educational information, not clinical or legal advice. The research on griefbots is limited and evolving, and individual experiences differ. If grief feels overwhelming, a counselor can help; if you are in crisis or thinking about harming yourself, call or text 988 or call 911.
Sources
- University of Cambridge - Call for safeguards to prevent unwanted 'hauntings' by AI chatbots of dead loved ones (May 2024)(Academic)
- Hollanek and Nowaczyk-Basińska - Griefbots, Deadbots, Postmortem Avatars: On Responsible Applications of Generative AI in the Digital Afterlife Industry, Philosophy & Technology 37:63 (2024)(Academic)
- Artificial Afterlife: Philosophical Reflections on Griefbots (Springer chapter, 2025)(Academic)
- Griefbots: A New Virtual Reality for Our Youths? (Springer chapter; survey of generational attitudes)(Academic)
- Frontiers in Genetics - Non-death loss and GenAI: ethical reflections on griefbots of the living (2026; literature review of griefbots for bereavement after death)(Academic)
- Texas Estates Code Chapter 2001 - Texas Revised Uniform Fiduciary Access to Digital Assets Act(Government)
About this review
Everendium Editorial Team · last reviewed October 7, 2026.
Reviewed for general accuracy against the University of Cambridge's published research on 'griefbots' (Hollanek and Nowaczyk-Basińska, 2024), later academic commentary, and Texas digital-assets law as of October 2026. The research base is small and mostly conceptual, not clinical trials; this is not clinical advice. Recommended for review by a bereavement clinician before wide promotion
Related glossary terms
Related reading
Make this easier for your family.
Record what you've just read in your Final Day Plan, then download it as a PDF. Nothing is stored online.