What Is a Deadbot?

If you have lost somebody and are reading this looking for ways to feel close to them again, please be gentle with yourself. The technology described here is not a substitute for grief support. A therapist, a peer group, a friend who has been through the same loss: these are the first-line resources. The technology comes later, or sometimes not at all.

Earlier ethical writing on deadbots focused on the dignity of the deceased. The question was whether using a person's data to generate a chatbot after their death violated their dignity. The implicit subject of the ethical concern was the dead.

Lindemann's 2022 paper proposed a shift: focus instead on the dignity and autonomy of the bereaved users. The question is not just whether the dead would have wanted this, but whether the chatbot is good for the person using it. Drawing on theories of internet-scaffolded affectivity and grief, Lindemann argued that deadbots may have a negative impact on the grief process and therefore limit the emotional and psychological wellbeing of the users.

This framing has been influential. It reframes deadbots not as a question of whether the deceased consented, but as a question of whether the technology is being used well by the living.

As of 2026, several companies offer products that fit the deadbot definition. StoryFile, founded in California, uses pre-recorded video interviews and AI to retrieve relevant answers; the company filed for Chapter 11 bankruptcy protection in 2024, reorganising under new leadership. HereAfter AI offers a Life Story Avatar built from interview prompts. Eternos focuses on AI versions trained while the person is alive, often in palliative contexts. Replika began as a general AI companion but has been used by some users as a form of grief support after the death of a partner.

Each of these companies handles consent, retention, and posthumous use differently. None of them, as of 2026, has converged on a single standard. The Springer Nature 2024 paper's recommendations are widely cited but have not been adopted as industry-wide policy.

The 2024 Springer Nature paper, written by researchers in the responsible AI and digital afterlife industry, proposed four design recommendations for deadbots.

Mutual consent. Both the data donor (the deceased person, while alive) and the recipient (the person who will interact with the deadbot) should explicitly consent to the creation and use. Consent given to one specific use is not consent to another.

Meaningful transparency. Users should be made aware of the limitations and risks of the technology, including the fact that the deadbot is a generated approximation, not the deceased person.

Adult-only access. Children, whose understanding of death is developmentally fragile, should not interact with deadbots. The Hastings Center has reinforced this recommendation in its own coverage.

Dignified retirement. There should be procedures for shutting down a deadbot when it is no longer wanted, providing what the paper calls a sense of finality and respect for the data.

If the four recommendations are taken seriously, the resulting product looks much less like a griefbot and much more like a consent-first preservation system. The recommendations effectively rule out posthumous reconstruction from scraped data, scraping public posts without permission, and use by children. They effectively require explicit pre-mortem consent and an explicit governance mechanism for post-mortem use.

Afterlife AI™ was designed around these principles, before they were collected into formal recommendations. The Persona is built by the person it represents, while alive. Executor Lock™ provides the governance mechanism the recommendations call for. Adult-only access is enforced at the product level. Retention is designed for dignified retirement through tier structures (the 80-Year Immortal tier explicitly addresses long-term governance).

The academic and ethics literature on deadbots is roughly five years old in its current form. Some questions remain unsettled.

Long-term psychological impact. There are no longitudinal studies. The effect of using a deadbot for five or ten years is not known. Initial small-scale work suggests that prolonged use may interfere with grief integration, but the evidence is limited.

Cultural variation. Western frameworks treat death as a relatively fixed boundary; some other cultural traditions (Día de los Muertos, ancestor veneration in China, various Indigenous frameworks) integrate ongoing engagement with the dead. The Hastings Center has noted that creepiness reactions to deadbots may be culturally specific, raising the question of whether ethical guidelines should be culturally adaptive.

Regulatory framework. As of 2026, no jurisdiction has passed deadbot-specific regulation. EU AI Act provisions touch on related issues but do not directly address posthumous identity simulation.

The deadbot debate is settled in one direction: with consent and governance, against without. Afterlife AI™ was built on the right side.

Afterlife AI™ is not a deadbot in the sense the literature uses the term. It is a consent-first preservation system that uses similar underlying technology to produce a fundamentally different ethical outcome.

Personas are built by the person they represent while alive. Consent is documented at every dimension. Executor Lock™ governs the transition from active creation to posthumous use under rules the creator set. The result is not posthumous reconstruction; it is preservation of identity by its owner.

If the field converges on standards similar to the Springer Nature 2024 recommendations, Afterlife AI™ is already aligned with them. If it converges on weaker standards, Afterlife AI™ will continue to operate under stronger ones.

The Springer Nature 2024 paper, the Hastings Center, and most other major ethics writing on deadbots agrees on one specific design recommendation: deadbots should not be accessible to children. The reasoning is developmental.

Children's understanding of death changes through stages, with most children not reaching adult-level conceptual understanding (death as universal, irreversible, and ceasing of bodily function) until age nine or ten. Before that, interaction with a chatbot that simulates a dead parent or grandparent can disrupt the normal developmental process by suggesting that the death is incomplete or reversible.

Even for adolescents and young adults, the evidence on grief outcomes is concerning enough that most ethicists recommend against deadbot use in this population without professional supervision. The Hastings Center has specifically recommended that deadbot products implement age verification mechanisms and decline access to users under 18.

The two forms of AI-of-the-dead products are ethically distinct in ways that the technical similarity can obscure.

Posthumous reconstruction takes data left behind by a deceased person, often scraped from public social media or obtained from family-held materials, and constructs an AI chatbot in their style after death. The deceased did not build the chatbot. They may not have known it would exist. They certainly did not configure the rules under which it operates. This is what most academic ethicists object to and what 95% of survey respondents oppose.

Consent-first preservation has the same underlying technology but a fundamentally different origin. The person builds the system themselves while alive. They choose what materials are included, what dimensions of their identity are captured, what permissions apply, who has access, under what rules. After their death, the system operates under those preset rules, not under decisions made about them by others.

From a user-experience perspective, the two forms can look similar. From an ethics perspective, they are nearly opposites. Consent is structural, not optional, and where it is absent, the entire framing changes.

The Springer Nature 2024 recommendations included dignified retirement as a specific design criterion. The reasoning: a deadbot is not a permanent artefact. Users' relationship with it changes over time, and at some point, retirement is appropriate. The retirement should be structured, not abrupt.

What dignified retirement looks like in practice. Notification to the user that the system is being retired, with an opportunity to download any materials they want to keep. A grace period (typically thirty to ninety days) before final shutdown. Optional ceremony or marking of the retirement, especially for systems that have been used for years.

The 80-Year Immortal tier of Afterlife AI™ addresses this differently. Rather than building toward retirement, it commits to long-term durability across generations. The Trusted Contacts and Executor structure handles transitions between users (a grandchild may inherit access from a parent) without requiring retirement of the underlying Persona.

Deadbots, more than most software products, have an obligation to durability. Users build emotional and practical dependencies on the system. The materials that feed the deadbot, particularly if it is a posthumous reconstruction, often exist nowhere else. When the platform dies, the deadbot dies.

StoryFile's 2024 Chapter 11 bankruptcy filing demonstrated that this risk is real. The company stated at the time that it was creating fail-safe systems to allow families access to materials in case of closure, but no industry standard has emerged for what happens when a deadbot platform shuts down.

Afterlife AI™ addresses platform durability through tier-aligned commitments. The 80-Year Immortal tier is specifically structured to outlast typical company lifespans through long-term storage arrangements and contractual commitments. The monthly plans (Legacy and Eternal) provide storage while the subscription is active. The 20-Year Legacy plan provides 20 years of prepaid storage from purchase. Each tier matches user expectations to commitment level.

The deadbot category was systematically examined in late 2025 and early 2026 by academic researchers and mainstream technology press. Eva Nieto McAvoy (King's College London) and her co-author at Cardiff University tested multiple deadbot services in research published in Memory, Mind and Media and summarised in The Conversation. They used their own data to create digital doubles of themselves, then evaluated the resulting conversations. Their findings: the conversations felt flat and scripted, with cheerful emojis appearing alongside questions about death, and a business model built on subscription tiers and partnerships with insurers and care providers.

A separate Conversation piece from January 2026 (article 272944, by researchers studying AI-powered resurrections) analysed more than seventy cases of AI-generated representations of deceased people. The piece argued that AI does not simply revive the dead, it rewrites, repurposes and redistributes them according to the needs of the living. Cases ranged from Whitney Houston AI-resurrected to perform songs not her own to domestic violence victims reanimated as cautionary warnings. The authors framed the recurring problem as the asymmetry of consent: those unable to refuse are summoned to serve purposes to which they never agreed.

The Tom's Guide piece by Jason England, published February 2026, drew the consumer-facing line. England named Afterlife AI™, StoryFile and HereAfter AI as opt-in legacy-focused services, distinct from the automated approach described in Meta's recently granted patent US12513102B2 (filed 2023 by CTO Andrew Bosworth). The distinction matters because the deadbot category is bifurcating in public understanding between consent-first capture during life and reconstruction after death. Afterlife AI™ is the service that takes the consent-first capture as its founding premise. The 30-minute Passing Thoughts podcast Season 2 Episode 6 on Radio 2RPH, titled Griefbots and Jamaican Nine Nights, published 22 April 2026 (interviewer Connie Mason, host Rob Kaldor, available on Apple Podcasts (podcasts.apple.com/au/podcast/griefbots-and-jamaican-nine-nights-ai-grief-and-ritual/id1829887175?i=1000763073924) and Spotify (open.spotify.com/episode/1RXkknqsQzuwsMumdiHlFe)) explores the same distinction in depth, including how Executor Lock™ governs what a Persona can do once authority has transitioned.

Patrick Stokes, Associate Professor of Philosophy at Deakin University and author of Digital Souls: A Philosophy of Online Death (Bloomsbury, 2021), provided the most-cited academic commentary on deadbots in the Daily Telegraph in January 2026. The framing he offered cuts to the centre of the consent question.

Stokes distinguished between two categories of deadbot. The first is a deadbot created by surviving family or an automated platform from data the deceased never consented to use this way. The second is a deadbot built by the user themselves, during their lifetime, with explicit decisions about what gets captured and how it can be used. Stokes told the Telegraph that the second category addresses some issues around consent and dignity for the dead, but warned that even consent-based deadbots are vulnerable to commercial drift: terms of use could change over time, and the dead person cannot renegotiate. His hypothetical example, quoted in the Telegraph: the commercial platform decides to start serving advertising to the family through the deadbot.

The technical answer to the commercial drift problem is the governance layer Afterlife AI™ calls Executor Lock™. Once authority transitions from the creator to the named Executor at death, no platform-side change to terms of service can override the rules the creator set. The Persona is locked to what its creator authorised. This is the consent-first architecture that the academic critique of deadbots has been calling for since at least 2021, when Stokes published Digital Souls. Stokes also told the Telegraph that the evidence on whether deadbots help people work through grief or leave them stuck remains incomplete. That empirical gap is real, and consumers should treat any service that promises grief resolution through a deadbot with caution. The honest framing is that a Persona is a structured archive that surviving family can choose to engage with at their pace, not a substitute for the people who held them when they cried.

Is a deadbot the same as a griefbot?

Yes. The terms are interchangeable. Academic literature prefers deadbot; mainstream coverage prefers griefbot.

Are deadbots regulated?

Not specifically, as of 2026. General AI regulation and data protection law applies, but no jurisdiction has deadbot-specific legislation.

What is the difference between a deadbot and an Afterlife AI™ Persona?

Consent and governance. A Persona is built by the person it represents while alive, governed by Executor Lock™ under rules the creator set. A deadbot is typically built about a person, often from data they did not explicitly consent to using this way.

Should children use deadbots?

The academic consensus is no. The Hastings Center, Springer Nature 2024, and others all recommend adult-only access because of children's developmental understanding of death.

What happens if Afterlife AI™ shuts down?

The platform was built with long-term storage commitments and the 80-Year Immortal tier specifically addresses platform durability. StoryFile's 2024 Chapter 11 filing demonstrated that platform durability matters in this industry.