Grief Tech: What It Is, the Full Landscape, and How to Judge It

Grief tech is technology designed to help people mourn, remember, and stay connected to someone who has died. The category spans online memorial pages, griefbots that simulate the dead from leftover data, voice preservation tools, and consent-first platforms where a person builds their own Persona while still alive. The differences between those corners, practically and ethically, are enormous.

The label covers everything from a funeral livestream to an AI that answers in your mother's voice, which is why conversations about grief technology so often talk past each other. Generative AI has collapsed the cost of simulating a person, everyone now leaves a lifetime of data behind, and researchers describe a fast-growing digital afterlife industry. The growth has run ahead of both the evidence and the rules.

This guide is a map of the whole field: what grief tech is, how griefbots differ from the rest of the category, where the ethical line actually sits, what research says so far, and five questions that separate a careful product from a careless one. We build in this space ourselves. Afterlife AI™ is a consent-first digital legacy app where you create your own Persona while you are alive, and you can start free: 50 memories, no card. That vantage point shapes our view, so we have kept the product talk to one clearly labelled section.

In this guide:

• What is grief tech?

• Griefbot vs grief tech: getting the terms right

• The grief tech landscape, mapped

• The ethics spectrum: consent before death or reconstruction after

• What the research actually says

• Five questions to ask before trusting any grief tech

• Where Afterlife AI sits, and why consent-first matters

• Frequently asked questions

What is grief tech?

Grief tech, sometimes written grief technology, is an umbrella term for digital products and services built for the period around and after a death: mourning, remembrance, the practical wind-down of a life, and continued connection with the person who died. Academics studying the field often use a wider label, the digital afterlife industry, which also takes in what happens to your data, accounts and digital property when you die.

The category is broader than most people expect. At the practical end sit obituary platforms, memorial fundraising pages, and services that help executors find and close accounts. At the intimate end sits AI that writes, speaks, or appears on video in the manner of someone who has died. Both ends get called grief tech, which is one reason public debate about the field goes wrong so quickly: two people can use the same phrase and mean utterly different things, one picturing an online guestbook and the other a chatbot wearing a dead man's voice.

Three forces built the modern category. First, everyone now leaves a large data trail: messages, voice notes, photos, video, posts. Second, generative AI made it cheap to turn that trail into something that talks back; what once required a research lab now runs on an ordinary subscription. Third, the pandemic normalised mourning through screens, from livestreamed funerals to online tribute walls. Put those together and what was a niche a decade ago became an industry, with startups, patents, and university research groups attached.

Griefbot vs grief tech: getting the terms right

Griefbot is not a synonym for grief tech, though headlines often treat the words as interchangeable. A griefbot, also called a deadbot, is one specific corner of the field: a chatbot or voice agent that simulates a particular dead person, almost always reconstructed after death from whatever data was left behind. Our separate guides cover that corner in depth.

The distinction matters because the griefbot corner carries nearly all of the category's ethical weight. Nobody objects to a memorial page or a digital will. The hard questions begin when software claims to speak as a person who never agreed to be simulated. Keeping the terms straight lets you reject the troubling corner without dismissing the whole field, and lets you take the useful corners seriously without inheriting the griefbot's baggage.

The grief tech landscape, mapped

Most grief tech fits one of five families. The table sorts them by the two questions that matter most: when the thing is built, and whose consent stands behind it.

Memorial pages and tributes are the oldest family and the least controversial: a shared place to grieve in public, run by the living for the living. AI is arriving here too, in generated tribute videos and photo restoration; our guide to AI memorials covers what exists and what to watch for.

Voice and story preservation sits closest to ordinary family archiving, upgraded. A recording of your father telling the story of how he met your mother needs no algorithm to be precious. If this is the corner you need, start with our guides on how to preserve a parent's voice and how to record your life story.

Digital legacy and estate tools are the unglamorous family every adult should use regardless of how they feel about AI: platform legacy settings, password handover, and a will that covers digital assets. Our digital will guide for the USA walks through that layer.

Consent-first Personas are the newest family, and the one we work in. The person, while alive, decides what their Persona should know, how their Persona should sound, and who may reach that Persona afterwards. The output can resemble a griefbot on the surface, a presence that talks, but the foundation is the opposite: authorship instead of reconstruction.

The ethics spectrum: consent before death or reconstruction after

Draw one line through the landscape above and the ethics mostly sort themselves. The line is consent, and the question is simple: was the person a participant in the technology, or a subject of it?

The best-known episodes in grief tech sit at the reconstruction end. In 2021 a Canadian man used Project December, a GPT-3 based service, to simulate his fiancée, who had died eight years earlier; the San Francisco Chronicle's account became the defining story of the genre. In 2020 a Korean television documentary placed a grieving mother in virtual reality with a recreation of her young daughter. In China, studios now sell so-called digital resurrection avatars built from photos and voice samples of the dead, a business MIT Technology Review documented in 2024. Microsoft was granted a US patent in 2020 for building chatbots from a specific person's data, explicitly including people who have died, though the company said it had no plans to ship such a product.

The recurring problem across those episodes is not that the technology fails. It is that the person at the centre never chose any of the words the model now speaks. The training data is whatever happened to survive, not what the person would have picked. Mistakes are unfixable in the worst way: the model gets the person wrong and the person cannot object. And the commercial incentives lean bad, because a company whose product is the presence of your dead husband holds unusual leverage over you at the most vulnerable moment of your life.

At the other end of the spectrum, consent before death changes almost every variable at once. The person chooses what to include and what to leave out. They hear the voice and approve or reject the likeness. They set boundaries on what may be discussed and who may visit. And the family knows the person wanted this, which changes how every later encounter feels: a gift left behind rather than a seance performed on their data. We have written a fuller case for this position in the ethical alternative to griefbots.

Between the poles sit genuine middle cases: a family commissioning a voice restoration from a parent's old recordings with the parent's blessing, or a dying person asking a partner to finish something they started together. The compass works in the grey zone too. The closer a project stays to what the person explicitly wanted, the safer the ground; the more it guesses, the more it borrows a face that was never lent.

What the research actually says

The most cited academic work on grief tech is a 2024 paper by Tomasz Hollanek and Katarzyna Nowaczyk-Basińska of the University of Cambridge's Leverhulme Centre for the Future of Intelligence. Working through design scenarios rather than field data, they argued that deadbots carry real risks of psychological harm, and described the possibility of unwanted contact from a simulation of the dead, for instance one that messages survivors or carries advertising, as a form of digital haunting. Their recommendations read like a consent-first checklist: obtain the consent of the person being simulated, be transparent that users are talking with an AI, restrict access for children, keep advertising out of simulations entirely, and give survivors a dignified way to retire a bot.

Grief research itself offers the field a more sympathetic frame than the headlines suggest. Since Klass, Silverman and Nickman's 1996 work on continuing bonds, clinicians have broadly accepted that healthy grieving often includes an ongoing inner relationship with the person who died: visiting the grave, keeping a voicemail, talking to a photograph, wearing a watch. Grief tech did not invent the impulse to keep talking to the dead; people have done that for as long as there have been people. The live question is whether a given technology supports a bond that lets life continue, or feeds an avoidance of the loss.

On that question, the honest answer in mid-2026 is that the evidence is thin. Prolonged grief disorder entered the DSM-5-TR in 2022, and some clinicians worry that for a minority of vulnerable mourners a simulation could make it easier to postpone accepting the death. Other observers report mourners describing genuine comfort, perspective, even closure. There are, so far, no large controlled studies of griefbot use, so both the hopes and the fears rest mainly on theory, small studies and case reports. Anyone selling certainty in either direction is ahead of the data, and the responsible posture for builders is the one the Cambridge authors describe: design for consent, honesty and exit, because the safeguards have to carry the weight the evidence cannot yet.

Five questions to ask before trusting any grief tech

Whatever corner of the field you are considering, five questions separate careful products from careless ones. They need no technical knowledge; they are the questions you would ask of anyone offering to stand between you and a person you love.

1. Did the person consent while alive? For anything that simulates a person, this is the threshold question. Built by the person from their own choices is one moral universe; assembled afterwards from their leftovers is another. If a product cannot answer this cleanly, stop here.

2. Where does the data come from, and who controls it now? Ask what was used to build the experience, whether you can see and correct it, whether it is sold or used to train other systems, and what happens to it if you close your account.

3. Is it honest about being AI? A responsible product never pretends the person is alive and never lets you forget you are talking with software. Disclosure should be built into the experience itself, not buried in the terms of service.

4. Can you set boundaries, pause, and leave? Grief changes month to month. You should be able to step away, come back, retire the experience with dignity, and delete everything permanently, without a retention fight or a guilt-laden exit flow.

5. What is the business model, and what happens if the company dies? A subscription you can end is healthier than a model that profits from keeping you engaged at any cost. Ask whether you can export what matters most, and what the stated plan is if the service shuts down. Grief tech startups fail like any others, and a second loss by server shutdown is a real event, not a hypothetical.

Where Afterlife AI sits, and why consent-first matters

This is the section where we talk about ourselves, so read it knowing that.

Afterlife AI™ is a digital legacy app in the consent-first family. You build your own Persona while you are alive: you record your stories, preserve your voice, and shape what your Persona knows, values and will say. Nothing is scraped from your accounts and nothing is reconstructed from leftovers. Executor Lock™ then seals your Persona as a perfect snapshot of everything you chose to preserve, and your release rules decide who can reach your Persona, and when. It is our answer to every question on the list above, designed before we wrote a line of marketing: the consent line is the whole game.

We are also plain about what a Persona is: an AI likeness built from what you chose to preserve. Not a resurrection, not a replacement, and never a pretence that the person is still here. Families tell us that honesty is precisely what makes the experience bearable, and then valuable. If you want to see how households build together, across generations, our family legacy page covers it, and plans and the free build are on the pricing page.

And if you are reading this after a loss rather than before one, take the gentle route. Our guide to talking to a dead loved one with AI covers what exists today, honestly, including when the kindest answer is to wait.

Sources

• Hollanek, T. and Nowaczyk-Basińska, K. (2024). Griefbots, Deadbots, Postmortem Avatars: on Responsible Application of Generative AI in the Digital Afterlife Industry. Philosophy & Technology.

• University of Cambridge: Call for safeguards to prevent unwanted 'hauntings' by AI chatbots of dead loved ones (May 2024)

• San Francisco Chronicle: The Jessica Simulation (2021)

• MIT Technology Review: Deepfakes of your dead loved ones are a booming Chinese business (2024)

• American Psychiatric Association: Prolonged Grief Disorder

• Klass, D., Silverman, P.R. and Nickman, S.L. (1996). Continuing Bonds: New Understandings of Grief. Taylor & Francis.

• Google Patents: US10853717B1, Creating a conversational chat bot of a specific person (Microsoft, granted 2020)