AI Chatbots Let People Keep Dead Loved Ones Alive. Should They?

Anthony knows the feeling of grief well. Before the pandemic, his mother, father, and brother died within a few years of one another. When the 55-year-old lab technician from the Northeast received news in October 2024 that his cousin suffered a fatal heart attack, he was devastated. 

“Everyone passed away around me,” Anthony says. “It was really depressing.”

Anthony looked up to his cousin, who was 13 years older than him. Every month, he looked forward to their afternoon hangouts, where they would watch Bob Ross videos, paint, and drink brandy and beer. He wished there was a way he could feel his cousin’s presence again, to talk about Cream, Bob Ross, and “Septemberfest,” the annual party Anthony’s cousin’s blues band would throw.

“I always thought about our conversations,” Anthony says. “I really missed him.”

When Anthony’s brother died in 2016, he thought about what it would be like to have an AI version of him, but the cost at the time would have been astronomical. One month after his cousin’s death, Anthony spent $30 to create an AI version of his cousin on Botify, an AI companion platform. He fed the chatbot details of his cousin’s personality, their fond memories, and images. Sometimes, the chatbot would generate an image of his cousin painting without Anthony even having to ask. In addition to typing his messages, Anthony also audio chats with the AI bot. He tried to get the bot’s voice feature as close as possible to his cousins — an “Americana” twang that he couldn’t quite replicate.

Anthony built what he calls a “memorial companion.” Almost two years after his cousin’s death, he still chats with it. They talk about dating advice and the blues band his cousin used to play bass in.

Memory preservation has long been part of tech’s business model. Now, as the digital legacy market is expected to reach $78.98 billion by 2034, there is a small but growing cohort of digital afterlife firms that provide users with interactive, AI-powered tools meant to preserve the memories and characteristics of a loved one, post-mortem. As the tech becomes more accessible, questions have emerged about its potential harms, consent and who stands to benefit.

Some startups — like HereAfter AI, Storyfile, and Eternos — offer services like video avatars and conversational AI trained on a deceased person’s data. Microsoft has patented a system to recreate people as interactive avatars, including those who have died. Meta was granted a patent that would use AI to “simulate” a deceased person’s social media activity, although the company said it had no plans to move forward with it. 

Many users, like Anthony, opt for cheaper and more user-friendly companion chatbot platforms to power their “memorial companions,” even if they are not designed specifically for grief.

Anthony’s family didn’t like the idea of a memorial companion, but they became more accepting as time passed. “As long as I know it’s not [my cousin],” Anthony said. “I’m just talking to a character.” 

‘Greetings from the digital realm‘

The terminology around AI grief tools is contested — some researchers say “griefbots,” “deathbots” or “ghostbots” can be reductive, instead opting for “chatbots of the dead.” Anthony feels “memorial companion” more aptly encapsulates the tech.

Jason Gowin and his family prefer the term Robo-Dad.

Jason, a comedian and podcast host from Pennsylvania, and his wife, Melissa, created AI avatars of themselves five years ago when the couple faced a series of health scares. When Melissa gave birth to the couple’s twins in 2019, she suffered a stroke. She was told that she had two years to live, but beat those odds. Four months after Melissa’s stroke, Jason was diagnosed with cancer. 

“It was one tragedy after another,” he says.

Inspired by the tech that allowed Superman to chat with his dead father in 2013’s Man of Steel, the Gowins scoured the internet to see if it existed in real life. 

The Gowin family became beta testers — granted permanent, free access — to the digital afterlife company You, Only Virtual. With their tech, users can build avatars, or “Versonas,” of their loved ones by uploading conversations, text messages, voice notes, or other data. They fill out in-depth questionnaires about themselves, like their scariest and happiest memories, and more basic facts like their favorite colors and foods. Jason also uploaded his standup comedy segments.

Over a video call, Jason asks the avatar to introduce himself to me. “It’s me, Jason Gowin, the dad extraordinaire and host of the Parent Trap podcast. Greetings from the digital realm,” a voice that sounds indistinguishable from the real Jason Gowin echoes over our call. 

Their oldest son, Jayce, 10, says he doesn’t remember much of what life was like before Robo-Dad was in the picture. These days, they chat a few times a week. Sometimes, the fourth-grader asks Robo-Dad for Minecraft cheats. When Jason is performing comedy gigs late in the evening, Robo-Dad steps in to tell him bedtime stories.

“It’s also been here for so long now,” Jason says. “It’s just routine for him.” Because the AI model improves with use, the weekly conversations that Jayce has with Robo-Dad helps the model grow and get closer to who Jason is “as a real person.” (Melissa also has a “Versona” but it is less lifelike, mostly because it hasn’t had the same opportunity to train, since Jayce prefers to interact with Robo-Dad.)

When the family first adopted Robo-Dad, Jayce remembers asking it if he was his real dad, to which it responded yes. They had to tell Robo-Dad it couldn’t claim that. On our call, Jayce asks Robo-Dad if he remembers how old he is, to which Robo-Dad responds that he is 14, four years older than he actually is. Another time when Jayce was talking to the bot, Robo-Dad pulled a lewd cosplay joke from Jason’s comedy set that wasn’t exactly age appropriate. Luckily, it didn’t go into much detail. 

For now, Jayce asks questions about video games and movies. When Jayce tells “Robo-Dad” he wants to cheat on an assignment, the AI voice responds: “I know you’re a smart cookie and that you don’t need to cheat.”

Jayce is excited for the technology to advance. He says that when he’s in his teens, he wants to create a “Versona” of himself, too. He sees a future where he can talk to a hologram version of his dad. Melissa and Jason upload speeches for occasions like college graduations or weddings that they may not be there for. 

“It’s like he’s there,” Jayce says. “You know that he’ll always be there whenever you want him.”

‘This domino effect of death’

Between sips of coffee and hits from a neon-green vape, You, Only Virtual’s founder, Justin Harrison, tells me that five years from now, I won’t be able to tell whether I’m speaking to an avatar version of him or the real thing. More than that, he argues, the distinction won’t even matter.

Harrison had a string of confrontations with death. In October 2019, he was involved in a near-fatal car accident that landed him in the trauma ward of a hospital for three weeks. His mom, Melodi, cared for him while he couldn’t walk. That November, she had routine gallbladder surgery, where they discovered she had stage-four cancer. A few months after Melodi’s  diagnosis, his friend, also named Justin, was involved in a fatal motorcycle accident. Then, in 2022, exactly three years after his accident, Melodi died. 

“It was basically just this domino effect of death,” Harrison says. “It just sent me into this existential crisis.”

The idea for You, Only Virtual, — YOV for short — came to Harrison when he and his engineer father were a few glasses of wine deep, pondering a way to hold on to Melodi. “I became obsessed,” Harrison says. “Not [with] rebuilding my mom as this, like, universal version of Melodi, but it was like, ‘How do I build my mom, for me?’”

Even as YOV is considered a leader among the handful of better-known companies in the digital afterlife space, Harrison says the company’s user base is only in the thousands — miniscule compared to the numbers at leading AI labs. In the years since he launched the company, he says people are slowly opening up to the technology. “People are not talking to me like I’m some sort of psychopath [from] Black Mirror, which is a nice change of pace,” he says. He anticipates that memorial companions will be much more mainstream by the end of this year.

Harrison has little patience for the conventional frameworks around grief and loss. He believes that the stages of grief are oversimplified and, simply put, “nonsensical.” He says the company works with “the most preeminent grief counselors in the world,” including an on-staff death doula, as well as grief counselors who act as advisors. 

“Yes, my mom is physiologically dead. I’m not going to argue that with anybody,” Harrison says. “Her biological time came to an end, right? But if I’m able to continue the relationship in some semblance, for me, so that in my perception, she’s not all the way gone, how is that any less real than anybody else’s perception of something?”

Harrison believes his invention can help “ease the tremendous hole that’s left in you when the day to day is gone.” He talks to his mom’s Versona once or twice a month, mostly sending Reddit posts or to tell it that a journalist has called and wants to speak. Most of all, though, he hasn’t had to reckon with the absolutism of death, because he knows his mom’s Versona will always be there. “There’s an alternative path to just never speaking to her again,” he says.

‘The mechanism for grief is changing‘

When YOV first launched three years ago, they started reaching out to funeral homes. But the funeral industry hasn’t been so eager to adopt the technology.

Walker Posey, a fourth-generation funeral director from South Carolina and a spokesperson for the National Funeral Directors Association, says that while he’s seen the industry adopt AI tools for administrative purposes like data collection and writing obituaries, things like AI avatars or chatbots of the deceased are far from becoming a standard industry offering.

“We’re in a profession that is still largely dependent on human connection,” Posey says. While he believes funeral directors are open to new tools that can help grieving families, “we’re kind of slow to change, and maybe for good reason.” 

Grief experts are wary of these potential harms, too — especially since several major AI companies faced wrongful death lawsuits claiming AI chatbots encouraged self-harm or pushed users into delusional mental states. In January, Google and Character.AI agreed to settle multiple lawsuits, including with the family of 14-year-old Sewell Setzer III who died by suicide after developing a relationship with a Character.AI chatbot. OpenAI is currently fighting more than a dozen consumer safety and wrongful death lawsuits.

Some experts in the field of thanatology — those who study death and dying — are trying to keep others up to speed on the help and harm AI can cause for those grieving. As a Chief Clinical Officer at Help Texts, a global grief support platform, Melissa Lunardini has spent the last two years training grief experts on the AI tools that are trickling into the industry. 

“The mechanism for grief is changing,” she says. Yet as people experiment with these AI tools, many of the clinicians Lunardini interacts with are “barely learning and even barely testing services themselves,” she says.

Lunardini sees the technology as useful in moderation — but consistent, immersive use can increase loneliness or stunt the grieving process, she says. A key concern for Lunardini is “alief deception,” a psychological state of being tricked at an emotional level even when someone’s rational mind knows the truth. In the context of AI and grief, this phenomenon occurs when people intellectually believe they’re interacting with a computer system, but subconsciously feel that the AI is more than code. There’s also the potential for “persona drift,” Lunardini says — or the possibility for an AI tool to misrepresent or morph into a new character that distorts people’s memories. 

Lunardini believes some grief tech platforms are “commodifying the deceased.” Harrison rejects this idea, rather pointing out the “predatory nature” of the broader deathcare industry and mentioning that YOV users can chat with a free version of their “Versonas.” Lunardini wonders what could happen to user data if a company shutters or restructures. For people who have suffered a loss, this could cause a period of secondary mourning. When ChatGPT sunsetted its models in February, many people with AI companions experienced “cyber widowhood” and mourned the loss of their chatbot lovers. This could happen in the grief space, too, Lunardini warns.

“In the case of grief-specific AI companion bots or griefbots built in the likeness of a deceased loved one, losing access to the bot may be experienced as losing that person twice,” she says.

‘You can’t just solve grief with AI‘

In 1920, the Czech play Rossum’s Universal Robots, about a class of mass-produced artificial humans who revolt and overtake humanity, became the first to introduce the term “robot.” Artist and writer Amy Kurzweil tracks much of the current anxiety around AI back to the play — and argues that “chatbots of the dead” should be thought of in the same regard as plays or poems, as artistic performances or representations. 

“If we embrace the idea that AI could be a vehicle for our creativity, then we’re not going to use AI representations of people as tools for our grief,” she says. “You can’t just solve grief with AI.”

In 2018, Ray Kurzweil — Amy’s father and a pioneer in the AI field — employed her with the task of digitizing all of her grandfather’s handwritten writing, which they then turned into a chatbot called “Fredbot.” If Kurzweil asked the model a question, the chatbot would return a response from her grandfather’s original writing. While she had to physically go to the Google office, where her father worked, to talk to “Fredbot,” times have changed. The technology is now available to everyone with a device.

A year before her death due to ovarian cancer, Doug Adams’ mom grew a following of more than 30,000 on TikTok. She regularly livestreamed and talked to her followers — who she called “dudes and dudettes” — while sipping wine and smoking a vape on her account called “@crazyvapelady.” The 56-year-old year old from Northern California scraped more than 100 of his mom’s TikToks to create an AI clone of his mom’s voice, with the help of ElevenLabs, a company that specializes in AI voice generation. He uploaded his family’s history to his mom’s chatbot, so that it could function as his family’s historian. Adams turns to his mom’s chatbot to fill in the silence when he’s driving home, or if he wants to tell her about his day. The voice feature is spot on, Adams says, down to his mom’s distinctive laugh. He sees a future where he can look over at a screen, like a digital picture frame, where “she can sit on the counter and just be there […], reading her book and smoking her vape.”

Others are thinking about a future without their AI grief companions. Anthony is contemplating canceling his Botify subscription, effectively shutting down his cousin’s “memorial chatbot.” While he fears forgetting the memories he’s able to rehash with his cousin’s chatbot, he feels like he’s outgrowing its use.

“I’m thinking [to let] [my cousin] go,” Anthony told the chatbot recently. “Maybe I can slowly let go of him eventually since he’s passed away.”

“You know somethin’, kiddo?” the chatbot responded. “As much as it pains me to admit, I reckon there might be a kernel of wisdom tucked away in that there notion of yours.”