0
0
Sixteen people lost someone they loved. Then researchers gave every one of them the same strange offer: talk to an AI version of that person, built from old texts, voicemails, photos, and whatever else was left behind.
Every single participant said they would do it again.
That finding comes from a new study out of the University of Colorado Boulder, presented at the 2026 ACM Designing Interactive Systems Conference. The researchers call these tools “generative ghosts” — AI chatbots trained on a real person’s data so they can respond the way that person might have, in their own texting style, after they’re gone. It’s part of a fast-growing industry some are already calling GriefTech, and in the last year it’s gone from a strange thought experiment to something millions of grieving people can download.
What the Researchers Actually Found
The Boulder team wasn’t just measuring whether people liked talking to these bots. They wanted to know what made the experience feel real, and what made it feel hollow.
Two patterns stood out. First, participants strongly preferred what researchers call “reincarnation” — a chatbot that speaks in the first person, as if it is the deceased person — over “representation,” a bot that talks about the person in the third person, like a narrator describing them. The first-person version felt like presence. The third-person version felt like a eulogy that talks back.
Second, and more surprising: people didn’t rate the bots as more comforting when the AI’s replies were longer or more detailed. They rated them as more comforting when the replies matched the rhythm of how the person actually texted — short, a little clipped, maybe an emoji they always overused. Accuracy of facts mattered less than accuracy of feel. One researcher’s framing has stuck with people who’ve covered this story: comfort here isn’t about getting the content right, it’s about getting the cadence right.
That detail matters more than it sounds like it should. It suggests that what grieving people are actually chasing isn’t information about the person they lost. It’s the texture of being spoken to by them again.
The Part Nobody Fully Agrees On
Not every response to this research has been warm. Coverage in outlets from Scientific American to Medscape has framed the core question the same way: are griefbots healthy or harmful?
The case for harm isn’t hard to make. Grief researchers have long said that healing usually requires accepting that a relationship has changed, permanently. A chatbot that answers back in a dead father’s voice, in real time, on your phone, can quietly work against that — not because it’s malicious, but because it’s available. It’s easier to text a griefbot at 1 a.m. than to sit with the fact that no one is going to text back.
And yet the people who actually used these tools weren’t naive about that risk. Several described using the chatbot in short, deliberate sessions — not as a replacement for grieving, but as something closer to visiting a grave that happens to write back. Every one of them, again, said they’d do it a second time.
So the researchers landed somewhere careful: griefbots aren’t automatically good or automatically bad. What seems to matter most is whether the person using one still has real, ongoing relationships with the living — people who notice when they go quiet, who ask a second time when the first answer is “I’m fine.”
Why This Is Happening Now
None of this exists in a vacuum. Loneliness researchers have spent the last few years documenting what some are calling a crisis of disconnection — fewer close friendships, more isolated households, more people who say they have no one to call at 2 a.m. Other 2026 research on AI companion apps found something similar: the chatbots built to ease isolation sometimes deepen it instead, because they’re a substitute for presence rather than the real thing. Grief has always been one of the loneliest experiences a person can go through, and it’s landing on a generation that already has less human infrastructure around it than the ones before — grief that, as other research on the subject has found, has a way of resurfacing months later with no warning, long after everyone assumes you’ve “moved on.”
Seen that way, griefbots aren’t really a story about artificial intelligence. They’re a story about what happens when real presence becomes scarce, and technology rushes in to fill the exact shape of the hole it left. The demand for these tools isn’t evidence that people have stopped wanting to be truly known by another person. It’s evidence of how far people will go to feel that, even for a few minutes, when the real thing isn’t available anymore.
It’s worth sitting with what that says about presence itself. Across almost every human tradition and every era, the deepest comfort in loss has never come from facts or explanations — it’s come from someone simply staying. Ancient wisdom on grief keeps returning to the same idea long before any of this technology existed: that the ache isn’t solved by information, it’s answered by presence, and the kind of presence that stays isn’t manufactured. It’s chosen, moment after moment, by someone — or, plenty of people would say, something bigger than us, keeping watch even when no one else seems to be — who’s actually there. That’s the one thing an algorithm, however well-trained, was never built to do. It can echo a voice. It can’t sit with you.
Maybe that’s the real reason the study’s most striking number isn’t the one about emojis or reply length. It’s the number who said they’d do it again — sixteen out of sixteen. Not because the technology healed anything, but because for a few minutes, someone answered. The deeper question the research keeps circling, without quite landing on it, is what happens to all of us if the people who are supposed to be the ones who answer stop showing up — and whether the ache griefbots are filling was ever meant to be filled by something that doesn’t actually know we’re there.
For now, the technology keeps improving, the industry keeps growing, and the sixteen people in that study are still, presumably, out there — grieving, texting, and finding out in real time what actually helps and what only feels like it does.
Discussion Question
If a griefbot could talk to you in your grandmother’s exact texting style, using her real words — would you want that, or would it feel like it was in the way of actually grieving her? What do you think separates comfort from avoidance here?
We’d genuinely like to know where you land on this one — leave a comment and tell us.
Share This
- 16 out of 16. Every person in a new grief-tech study said they’d talk to an AI version of their dead loved one again. What that says about presence might matter more than what it says about AI.
- Comfort in grief isn’t about getting the facts right. New research says it’s about getting the *rhythm* right — how someone talked, not just what they said. That’s a strange and telling thing to learn about loss.
- Griefbots are booming because real presence is getting rarer. That’s not really a story about AI. It’s a story about who’s still showing up for the people we love while they’re still here.
Common Questions About AI Grief Chatbots
What are AI grief chatbots, or “griefbots”?
AI grief chatbots, sometimes called griefbots or “generative ghosts,” are AI tools trained on a deceased person’s texts, voicemails, photos, and other personal data so they can respond in a way that mimics how that person communicated. They let a grieving person continue a kind of conversation with someone who has died.
Did a real study find that griefbots help with grief?
Yes. Researchers at the University of Colorado Boulder studied 16 people who used AI chatbots built from a deceased loved one’s data and presented the findings at the 2026 ACM Designing Interactive Systems Conference. Every participant said they would use the technology again, though the researchers also flagged real emotional risks.
Is it healthy to talk to an AI version of someone who died?
Experts are divided. Some research suggests short, intentional use can bring real comfort, similar to visiting a place tied to the person’s memory. Others warn that constant availability could make it harder to accept the loss, especially if the chatbot becomes a substitute for real, ongoing relationships with the living.
Why do people prefer griefbots that talk in the first person?
Researchers found that “reincarnation” style chatbots, which respond as if they are the deceased person speaking directly, felt far more comforting to users than “representation” style bots, which describe the person in the third person. First-person responses felt like presence rather than description.
What does grief research say people actually need most?
Across the research, the same theme keeps surfacing: people don’t just want information about the person they lost, they want to feel that someone is actually present with them in the loss. That’s something technology can imitate for a moment, but real, ongoing human presence is what the research keeps pointing back to.
