When AI Enters the Grief Journey

When I joined Roberta’s House as its executive director, I spent much of my first six months experiencing our programs and services firsthand. I sat in grief groups and listened as people talked about the hardest parts of life after the death of their significant person. I was especially interested in what they missed most about the ones they had lost. Often, it was something wonderfully ordinary: hearing that person’s voice from another room, watching them dance while cooking, or being greeted at the door after a long day.

Families have always found creative ways to hold on to what they remember most about people who have died. They save voicemails, wear memorial T-shirts, celebrate birthdays, display photographs, and keep ashes close. As technology changes how we capture and share memories, AI is offering families new ways to remember loved ones and feel connected to them. Some tools use a person’s recordings to synthesize their voice, allowing family members to hear new words spoken in a voice they know by heart. Others draw on texts, emails, and stories to create chatbots that can carry on a conversation in a style meant to resemble the person who died. Digital avatars add a visual presence, pairing a generated voice with an image that appears to look at and speak to the viewer. Together, these tools can make an interaction feel strikingly familiar, even though the words and responses are generated by AI. Researchers sometimes call these tools griefbots or digital recreations.

These technologies can feel like an extension of the ways families already preserve memories, but they introduce a meaningful difference. In the sections that follow, I’ll look at the tools available to families, how they differ from more familiar forms of remembrance, and how we as grief professionals can help people understand their options and make informed choices before using them.

What kinds of AI grief tools are available?

People have long turned to journals, support groups, and trusted listeners to make sense of a loss. Some are now bringing those same thoughts to general-purpose AI tools such as ChatGPT, drawn in part by the ability to receive an immediate response at any hour. That use is worth understanding, even as questions remain about what these tools can reliably offer someone who is grieving.

AI tools designed specifically for grief support, however, take this approach a step further.

Like a workbook or guided reflection exercise, they may offer prompts, information about grief, or coping strategies. Some follow a structured program; others use AI to respond to what someone shares in their own words. That back-and-forth can make support feel more personal than reading a page of advice. It also means families should know how the responses are generated, whether grief and bereavement professionals helped design the tool, and whether a live, qualified person is available when more support is needed. These features can vary by both product and subscription level.

Other AI tools like digital recreations build on another familiar practice: saving the traces of someone we love.

While voicemail preserves words a person spoke, voice synthesis technologies can use those recordings to produce new words in a voice that sounds eerily similar to theirs. Texts, emails, and stories can similarly be used to create a chatbot that replies in a style meant to resemble the person who died. Digital avatars can be used as images that appear to move and speak. These tools can offer a sense of interaction, but the new responses are generated by AI; they are not memories of something the person said or did. Researchers sometimes call these interactive recreations griefbots or postmortem avatars.

All of these tools grow out of recognizable needs: to be heard, to remember, and to stay connected. What changes is that AI responds and creates something new. Understanding that difference can help families decide what feels meaningful to them and help grief professionals support those decisions with care.

Where Caution Matters

The appeal of these tools is easy to understand. When someone longs to hear a familiar voice or share one more thought with a person they love, a digital recreation may offer a moment of comfort. We can respect that longing while helping families consider what may happen during and after the interaction. That conversation begins with the possible effects on grief, then extends to questions about consent, privacy, and control.

One clinical concern is that a recreation can sound convincing while generating words the person who died never said. If it says “I forgive you,” offers advice about a family conflict, or describes an event that never happened, those words may carry emotional weight beyond an ordinary chatbot response. A person may find the exchange comforting, distressing, or both. Repeated conversations could also make it harder for some people to distinguish their own memories from what the system has generated or to step away from an interaction they have come to rely on. These are potential risks, yet we do not have strong long-term evidence about how using digital recreations affects bereavement.

Children may need particular support in making sense of a lifelike recreation. An adult can explain that the voice or image was generated, but a child’s understanding of that explanation may differ from their emotional experience of hearing or seeing it. Families should consider the child’s age, whether the child wants to participate, and how they respond afterward. A trusted adult should be available to listen and help them understand what happened.

The practical questions begin with the material used to build the recreation. Recordings, texts, and photographs can include private information about other people who never agreed to its use. Family members may disagree about whether their loved one would have wanted to be recreated or who should have access. They should also understand the company’s terms: what happens to the information they upload, whether it can be deleted or downloaded, and what happens if prices rise, the product changes, or the service closes. For someone who has grown attached to the recreation, losing access could bring another painful sense of absence.

Equity deserves attention, too. A systematic review found that AI applications can create unequal opportunities across racial groups, and a review of AI in health care explains how data that leave groups out or misrepresent them can reinforce bias. Neither study evaluated AI grief tools, so these findings do not establish how often such tools fail grieving families. They do give us reason to ask whose voices, language, and experiences shaped a tool and whether it has been evaluated with the communities it aims to serve. A response that overlooks a family’s cultural practices or circumstances may feel less helpful, even when it sounds fluent.¹ ²

How Bereavement Professionals Can Help

Grief professionals do not need to decide for families whether an AI tool belongs in their grief journey. We can begin by listening: What does the person hope the tool will offer? Are they looking for a way to preserve memories, a place to talk about their loss, or the feeling of speaking with their loved one again? Those are different needs, and understanding the need helps us have a more useful conversation about the tool.

We can also give families language for what they are experiencing. The idea of continuing bonds, for example, recognizes that a relationship with someone who has died can remain meaningful. The dual process model describes how people move between attending to their loss and managing the demands of life after it. Neither theory tells a family whether to use AI. Both can help explain why someone might want to hear a familiar voice on one day and need distance from it on another.

From there, we can help people make a grief support plan that gives them more than one place to turn. Depending on their needs, that might include individual therapy, a peer support group, conversations with family or friends, self-help guides, journaling, remembrance rituals, and coping strategies for difficult moments. If someone is considering an AI tool, we can talk through where it might fit in that plan, what they would do if an interaction became upsetting, and who they could contact for human support. We can also help them recognize when grief is seriously disrupting daily life and a clinical assessment may be appropriate.

Finally, we can help families ask informed questions before they sign up or upload anything: Is this tool offering grief guidance or recreating a loved one? What information does it collect, and can the family delete it? Did the person who died consent to a recreation? Who can access it? Is there human support if the conversation becomes distressing? When children are involved, adults should consider the child’s understanding and wishes as well as how they will support the child afterward. Our role is to make room for both the longing behind these choices and the practical information families need to make them. We can also ask who tested the tool, whether its responses were assessed across diverse communities, and how users can report an inaccurate or culturally insensitive response.

Resources and Support

Roberta’s House grief support programs — programs for grieving children, adults, and families.

APA Dictionary: Continuing bond — a brief explanation of maintaining a connection with someone who has died.

The dual process model of coping with bereavement — foundational research on moving between attending to a loss and adapting to daily life.

American Psychiatric Association: Prolonged grief disorder — information on when grief may call for clinical assessment.

FTC: Using a health app? — questions to ask about an app’s use of personal information.

Griefbots, Deadbots, Postmortem Avatars — scholarly analysis of consent and responsible design of digital recreations.

Intahchomphoo C, Gundersen OE. “Artificial Intelligence and Race: a Systematic Review.” Legal Information Management. 2020;20(2):74–84. doi:10.1017/S1472669620000183

Norori N, Hu Q, Aellen FM, Faraci FD, Tzovara A. “Addressing bias in big data and AI for health care: A call for open science.” Patterns. 2021;2(10):100347. doi:10.1016/j.patter.2021.100347