Opinion | AI can resurrect the dead. Ask my chatbot mom.
By Gammatek ISPL Last updated: September 2026 | 14 min read
Author note: This piece draws on reporting and research published across multiple outlets covering AI "griefbot" or "deadbot" technology, cited throughout. It does not reproduce any single article's analysis or language — all framing and argument below is original.
Why This Matters
A small but fast-growing number of companies now let you build a chatbot that talks like a specific dead person — using their old texts, voice recordings, emails, and social media posts to recreate how they spoke, what they believed, and how they responded to things. Some of these "griefbots" can hold ongoing conversations for years. Some can be connected to the internet so they stay current on world events. If you've lost someone, or will, this isn't a hypothetical future technology — it's something you can sign up for today, and the research on what it actually does to the grieving process is still being written in real time, which means you're being asked to make a decision most psychologists haven't finished studying yet.
What These Tools Actually Are
The category goes by several names — griefbots, deadbots, grief tech — but the mechanism is consistent: a company takes whatever digital trace a person left behind (text messages, emails, voice memos, social media history, sometimes video) and uses it to train a chatbot, voice clone, or in some cases a video avatar that responds the way that person likely would have.
Some services stop there — a static recreation frozen at the point the data was collected. Others, notably the platform You, Only Virtual, let the AI avatar continue accessing the internet and "grow" through ongoing conversation, meaning the digital version of a deceased person keeps developing a personality independent of who they actually were when alive (source: GIGAZINE, reporting on AI ethics researcher James Muldoon's analysis).
The use cases people report range widely. Some talk to a parent's griefbot a few times a month, finding comfort just in knowing the option exists (source: Washington University's Common Reader magazine, reporting on a user named Harrison and his "Dadbot"). Others, like a Silicon Intelligence executive who built a chatbot of his late mother, describe the experience in much more totalizing terms — one report quoted him saying that whether his mother is alive or dead stops mattering once he can still find her and talk to her through the bot (source: NPR, via TechRadar).
This Isn't New — It's Just Newly Accessible
Fictional and speculative versions of this idea predate the current wave of AI products by years — the Black Mirror episode "Be Right Back" imagined a woman recreating her dead boyfriend through his digital footprint back in 2013, and it's frequently cited by both researchers and journalists covering this space as the cultural reference point everyone reaches for (source: The Algorithmic Bridge). What's changed isn't the concept — it's that building a conversational, voice-matched version of a specific dead person used to require custom engineering and now takes a consumer-facing app and an afternoon of uploading old messages.
One of the earlier real-world examples: a man named Joshua used a tool called Project December, built on GPT-3, to recreate conversations with his late fiancée Jessica, customizing the chatbot with just a short sample of how she talked and a brief description of their relationship (source: The Algorithmic Bridge). That story, once unusual enough to make international news, is now a fairly ordinary use case across several competing platforms.
A Comparison: How Current Griefbot Approaches Differ
Not all of these tools work the same way, and the differences matter for anyone considering one. Here's a structural comparison of the main approaches currently in use, based on how they're described in available reporting:
Approach | Data Source | Can It "Grow" Over Time? | Primary Format | Typical Context |
Static text chatbot | Uploaded texts/emails | No — frozen at creation | Text conversation | Personal, private use |
Internet-connected chatbot | Uploaded texts/emails + live internet access | Yes — develops independently | Text conversation | Personal, ongoing relationship-style use |
Voice-only recreation | Voice recordings | No | Audio playback/limited conversation | Memorial services, funerals |
Video avatar / VR recreation | Photos, video, sometimes VR capture | Varies by platform | Visual + audio interaction | High-emotion, often one-time reunions (e.g., VR) |
Public "ghostbot" of historical or public figures | Public record, writings | Varies | Text/voice | Entertainment, education — not personal grief use |
The "internet-connected, continues to grow" category is the one generating the most ethical debate, because it means the chatbot stops being a recreation and starts being something new — a system that will, over years, say things the real person never said and likely never would have.
What the Early Research Actually Shows
This is where most popular coverage of griefbots gets thin, so it's worth being specific about what's actually been studied versus what's still speculation.
The psychological case for caution is fairly consistent across experts who've weighed in. A clinical perspective published in a reader-response piece on AI and ethical dilemmas argued directly that griefbots are likely to prolong the effects of loss rather than resolve them, on the reasoning that healthy grief requires accepting an ending, and a chatbot that keeps responding removes that ending (source: WSJ reader-response series, via TOVIMA). The same piece raised a specific risk worth naming plainly: after enough time has passed, some users may begin to blur the line between "a chatbot built to sound like Mom" and actually feeling they are talking with their mother, even years after her death.
Psychologist Elaine Kasket, writing in the Oxford Handbook of Cyberpsychology, frames the stakes more broadly— arguing that the combination of AI and social media's permanence may be shifting not just how individuals grieve, but the broader cultural definition of death itself, and even how people think about their own mortality (source: The Globe and Mail).
On the other side, researchers studying actual users report a more mixed picture than the "this is obviously harmful" framing suggests. Reporting on griefbot platform Project December found that most users turning to these tools were already experiencing significant trauma and grief, and many described the chatbot as a coping mechanism rather than a delusion — suggesting the effect may depend heavily on how self-aware the user is about what they're interacting with (source: Al Jazeera).
A researcher at the University of Essex focused specifically on AI ethics has argued the decision shouldn't be left entirely to the companies building these tools — pointing out that griefbot platforms are overwhelmingly built and governed by startups and venture capital interests, not by grief researchers, ethicists, or the families who will actually use them, which raises a structural question independent of whether any individual user finds the experience helpful (source: GIGAZINE, citing James Muldoon).
The Question Nobody's Really Answered: What About the Dead Person?
Most coverage of this topic — understandably — centers on the living person's grief experience. A less-discussed angle, raised explicitly by at least one researcher covering the space, is the question of the deceased person's own rights: did they consent to being recreated this way? Most people who died before this technology existed obviously never had the chance to say yes or no (source: Common Reader, Washington University).
This isn't a hypothetical concern. Several platforms now offer "ghostbot" versions of historical and public figures who died long before AI chatbots existed — recreations of Confucius, of Holocaust survivors, even of Walmart founder Sam Walton, built explicitly for educational or entertainment purposes rather than personal grief (source: The Globe and Mail). None of them consented to this use of their likeness, because consent wasn't technologically relevant when they were alive. As the underlying technology gets cheaper and more accessible, the gap between "a grieving family member choosing this for themselves" and "a company building a public product out of someone's digital remains without their input" is likely to become a much bigger legal and ethical fault line than it currently is.
An Implementation Consideration for Anyone Weighing This Decision
If you're considering building or using a griefbot — your own or a family member's — a few concrete questions worth answering before you start, based on the patterns in the research above:
Decide in advance whether you want a static or growing chatbot. A frozen recreation based only on existing messages is a fundamentally different emotional object than one that will keep generating new material indefinitely — pick deliberately, not by default.
Set a check-in point with yourself, not just the technology. Several of the accounts above describe users who found comfort in occasional use but became uneasy with constant, ongoing conversation — deciding your own boundary ahead of time is easier than discovering it after the fact.
Consider whether the person would have wanted this, even informally — through something they said in life, a value they held about privacy or digital permanence, rather than assuming consent because they can no longer object.
If you're building this for a service offered to others (a funeral home, a memorial platform, a public figure's "ghostbot"), the consent question above isn't optional — it's close to the central ethical issue the available research keeps returning to.
Where This Is Headed
None of the researchers or reporters cited above are predicting this technology disappears — if anything, the opposite: it's getting cheaper, more realistic, and more normalized by the month, from funeral-home QR codes to full video avatars. One widely cited projection suggests that at current trends, the number of dead profiles on Facebook alone will outnumber the living by 2098 (source: The Globe and Mail) — and griefbots are a natural next layer on top of that existing mountain of digital remains.
The honest version of where this leaves us: griefbots aren't obviously good or obviously harmful as a category — the research so far suggests the outcome depends heavily on the specific design (static vs. growing), the specific user's psychological state going in, and whether anyone — the user, the platform, a grief counselor — is actively thinking about the boundary between comfort and avoidance. Right now, most of that thinking is being left to individual grieving people to figure out alone, with very little structural guidance from the platforms building these tools or the regulators who haven't caught up yet. https://www.gammateksolutions.com/post/it-s-all-fun-and-games-until-you-give-ai-your-credit-card
Comments