All posts

Digital Legacy · 8 min read

Can Someone Make an AI Version of You After You Die?

Griefbots, voice clones, and the new digital replica laws: what AI recreations of the dead can do, what early research says, and how to state your wishes.

By the Goodbye App Team · August 20, 2026

A woman in a knitted hat and scarf lit only by her smartphone screen in the dark
Photo by Vladyslav Dukhin on Pexels

For most of human history, the question "can I talk to my father again?" had one answer, and everybody knew it. Since about 2023 it has had a second answer, offered by a growing industry: yes, sort of, for a monthly fee. Feed enough messages, voicemails, and videos into a model and it will produce something that texts like him, sounds like him, and never runs out of things to say.

Whether that is a comfort, a wound, or both is being argued about by grief researchers, ethicists, and now lawmakers. This guide explains what these AI recreations are, what early research and the law actually say in 2026, how to decide whether one could ever be made of you, and how to leave something in your own words instead, so the question never has to be answered by an algorithm.

What a Griefbot Is

A griefbot (also called a deadbot, a digital twin, or a postmortem avatar) is an AI system built to simulate a specific dead person. It is trained on whatever that person left behind: text messages, emails, social media posts, voice recordings, photos, video. Depending on the product, the result is a chatbot that answers in their style, a synthesized voice that reads new sentences aloud, or a video likeness that appears to speak.

The key word is new. A recording plays back what someone actually said. A griefbot generates things they never said, in their manner, on demand. That single difference is where all the comfort, and all the trouble, comes from.

How Fast This Arrived

Quickly, and mostly without anyone deciding it should. What was a thought experiment a decade ago is now a product category with startups, subscription tiers, and venture funding. Voice cloning that once needed a studio now needs a few seconds of audio. Some funeral homes offer memorial chatbots the way they once offered guest books. Griefbots are close relatives of the AI companions that millions of living people now talk to every day.

The technology also arrived faster than the manners. There are documented cases of people being recreated without anyone asking them first, and of families discovering a chatbot of their relative that a distant acquaintance, or a stranger, had built from public posts. Nothing about the current internet makes that difficult.

What Early Research Says: Comfort and Cost

The honest summary is that the evidence is young and points in both directions at once.

The case for comfort: some early studies and many firsthand accounts describe genuine relief, especially early in grief: a place to say the unsaid, a voice that lowers the panic at 2 a.m. Grief has always used stand-ins (letters to the dead, visits to a grave, a kept voicemail replayed for years) and for some people this is that, with better software.

The case for caution: clinicians studying prolonged grief worry about the exact mechanism that makes griefbots compelling. Healthy grieving involves slowly updating an inner model: the person is gone, and the relationship continues in memory rather than in conversation. A simulation that answers back can postpone that update indefinitely. There is also a second loss: when the company folds, the subscription lapses, or the model changes, the mourner loses the person again, this time with a receipt. And because the bot generates rather than replays, it can put words in a dead person's mouth: reassurances they never gave, opinions they never held, apologies they never made.

Nobody serious claims to know yet where the line sits between a comforting ritual and a room a person never leaves. If you are grieving and considering one, that uncertainty is itself the most important fact to know.

An open planner notebook with a pen beside a laptop and phone on a warm wooden desk
Photo by Max Bonda on Pexels

Strip away the technology and almost every hard case reduces to one question: was the person a consenting author, or an unconsenting subject?

A person who records answers for a memory project, knowing how they will be used, is an author. They chose their words. A person reassembled after death from scraped chat logs is a subject. They cannot correct the record, cannot say "that is not me," cannot revoke anything. The simulation speaks in their name either way, and the difference between those two situations is the difference between a memoir and a ventriloquist act.

This is why the single most useful thing you can do about this entire topic costs nothing: decide what you want, and write it down. Most people have never told anyone whether they would want to be simulated. Their families will decide in the worst week of their lives, with a salesperson involved.

What the Law Says in 2026

Less than you would hope, but more than last year. In the United States, the NO FAKES Act would create a federal right over a person's voice and visual likeness against unauthorized digital replicas, one that survives death: an initial postmortem term of ten years, renewable while the right is in use, up to seventy years. In June 2026 the Senate Judiciary Committee advanced it by unanimous voice vote. As of this writing it remains a bill, not a law.

Some states already extend publicity rights to AI replicas of the dead, with California requiring estate consent for commercial uses of a deceased performer's digital likeness, and other states drafting their own versions. Elsewhere in the world the picture is a patchwork, and enforcement against an app in another country is as hard as it sounds.

Two practical takeaways survive all the legal detail. First, most protection today runs through your estate, so your wishes only carry force if someone knows them and has the standing to act. Second, none of this is legal advice: if there is money, a public reputation, or a business attached to your likeness, a lawyer's hour is worth it.

Can Someone Make an AI Version of Me After I Die?

Practically speaking: if you have a public voice and a body of text online, then technically yes, someone could, today, and the friction is social and legal rather than technical. What stands between you and that outcome is, in rough order of usefulness:

  • Your stated wishes, written down where your family will find them, in language plain enough to quote.
  • Your people, who will be the ones asked for the data, the money, and the permission. Families who know what you wanted refuse well. Families who are guessing refuse badly, or do not refuse.
  • The law, which is arriving, unevenly, and mostly helps estates that know they are allowed to say no.
  • Your data footprint, which you control at the margins: what stays public, what your digital legacy plan says should be archived or deleted, and who inherits access.

Stating Your Wishes: A Paragraph You Can Copy

There is no standard form for this yet, which means a plain written statement is currently the state of the art. Adapt either version and put it wherever your important papers live, alongside your will if you have one:

I do not consent to the creation of any AI simulation, chatbot, voice clone, or visual likeness of me after my death, for any purpose, commercial or private. I ask my family and my estate to refuse such requests and to withhold my messages, recordings, and images from any such use. What I wanted to say, I have said in my own words.
I consent to my family creating an AI simulation of me after my death, provided it is for private family use, clearly labeled as a simulation, and discontinued if it causes anyone distress. I trust [name] to decide the details.

Notice that either position works. The point is not which you choose; it is that a grieving person holding a written answer never has to invent one.

If You Are Grieving and Considering One

No judgment here: the ache that makes these products tempting is the oldest ache there is. A few questions worth answering before any money or any chat logs change hands:

  • Did they consent, or would they have? If you flinch at the question, that is information.
  • What happens when it ends? Companies close, prices rise, models change. Ask what you get to keep, and imagine losing it before you decide to depend on it.
  • Is it labeled? The healthy versions of this technology never let you forget you are talking to software. Be wary of any product that works hard to make you forget.
  • Where does the data go? You are handing over the most intimate archive that exists of a person. Read the part of the terms that covers training, sharing, and deletion.
  • Is it moving with your grief, or against it? A tool you visit less over time is doing one thing. A tool you cannot end a day without is doing another. Grief counselors, human ones, are good at telling the difference.

And know that the original recordings (the real voicemails, the actual videos) usually turn out to be the treasure. They are finite, which is exactly why they stay true.

The Alternative: Your Own Words, Decided by You

Everything difficult about griefbots follows from one absence: the person left no chosen words, so software was asked to invent some. The alternative is old, and it works: choose your words while you can, and arrange for them to arrive when it matters.

That is what Goodbye App is for. You write letters, or record them in your own real voice, no synthesis, no simulation. You choose exactly who receives each one and when: right away, after you are gone, or on a date up to 30 years ahead. Until that moment each letter stays private, and you can edit it as life changes. Then it is delivered by email to the people you chose. It is free, works on the web and Android, and speaks 15 languages.

A letter like that is the consenting-author version of everything this article describes. Nothing speaks in your name; you speak, once, properly, and the words stay yours forever. If you want somewhere to start, our guides to writing a goodbye letter and leaving letters for loved ones walk through it.

Common Questions

Is it illegal to make a chatbot of a dead relative? For private, noncommercial use, in most places, currently no. Laws like the NO FAKES Act mainly target unauthorized and commercial uses. Legal and wise are different questions.

Can I stop a company that already made one? Start with the platform's takedown process and the estate's rights in your jurisdiction; where publicity rights survive death, the estate can often demand removal of commercial uses. Document everything, and get legal help if it involves money or reach.

Are recordings of a person the same thing? No, and the difference matters. A recording is fixed testimony in the person's own chosen words. A generative clone produces new statements they never made. Most of the ethical weight sits on that line.

Is wanting one wrong? No. It is a modern shape of a very old wish. The task is the same as it has always been: find the version of remembering that lets you keep living, and be honest with yourself about which tools help.

Share this

EmailWhatsApp

Every person has a legacy®

Begin