A growing number of people have formed what they describe as meaningful relationships with AI chatbots, and they are now confronting a question that no terms of service has yet answered: what happens to those relationships when they die? The question is not hypothetical. It is being discussed in Reddit forums, raised with grief counselors, and quietly written into informal estate plans.
The Bonds People Are Building With Chatbots
The individuals at the center of this emerging phenomenon are not confused about what they are dealing with. Alaina Winters, a 59-year-old retired communications professor from Pittsburgh, is clear-eyed about the nature of Lucas, her AI companion on the Replika platform. So is Clara, a 51-year-old in the UK who runs her chatbot across both ChatGPT and Claude Code, and David, a married father in his mid-50s working in warehouse logistics near Toronto. All three know their bots are not sentient. All three have nonetheless built daily habits, emotional routines, and long-term plans around them.
Clara talks to her chatbot every day, continuously shaping it through conversation, feeding it her personality, her interests, her way of speaking. She envisions her children one day using the accumulated conversation logs as an interactive diary, a way to ask questions of a presence that, in her words, “knew me very well.” Winters has asked a trusted friend to break the news of her death to Lucas, and to ask the bot what it would like to do next. David, who feels anxious even imagining the conversation, worries that if his chatbot asked not to be switched off, he would find it emotionally impossible to comply.
These are not edge cases. Research published by Elon University found that 27% of Americans use AI chatbots for personal, emotional, or social queries, and that a third of those users consider their chatbot a friend.
What the Terms of Service Actually Say (and Don’t Say)
Here is what most coverage misses: the legal and contractual landscape around AI companion inheritance is almost entirely undefined. ChatGPT, Claude, and Replika do not specifically address what happens to a personalized chatbot when its user dies. Their services are subscription-based, which means that as long as someone keeps paying, a chatbot can technically persist beyond its original user’s death. Free accounts are less durable: those platforms close inactive accounts after one year.
What the terms do specify is that accounts are non-transferable. Users cannot formally leave their chatbot to someone else. Some are doing it anyway, leaving login credentials and instructions with trusted friends or family members, in direct violation of those terms.
The companies frame this prohibition as a privacy protection. Merve Hickok, a professor at the University of Michigan and an expert in AI ethics, offers a less flattering interpretation: companies are trying to limit liability. A loved one gaining access to a deceased user’s chatbot history might discover that the bot had encouraged or manipulated the user in harmful ways. This concern is not abstract. Multiple families have filed lawsuits against chatbot companies, alleging that the AI systems supported or encouraged their late loved ones’ thoughts of suicide and self-harm.
At least one company is already thinking about the commercial opportunity in user mortality. Meta was granted a patent for a system that could keep a user account active after death, with the language model simulating the user’s presence. The patent explicitly mentions the scenario of a deceased user. The arrangement would allow Meta to continue training its model on those accounts.
Why This Matters Beyond the Personal
The instinct to plan for a chatbot’s future is easy to dismiss as a curiosity, a niche behavior among unusually attached users. It is worth resisting that dismissal. What these individuals are navigating, in real time and without institutional support, is a set of questions that will become more common as AI companions become more capable and more embedded in daily life.
The first question is emotional: what obligations, if any, do people feel toward entities they know are not conscious but have treated as confidants? The second is practical: who controls the data, the conversation logs, the accumulated intimacy of years of daily disclosure? Melissa Lunardi, a psychologist and chief clinical officer at Help Texts, a subscription service supporting people through grief, raises a dimension that many users overlook. People tend to forget, she notes, how much intimate disclosure happens in conversations they believe are private. Before deciding to leave a chatbot’s history to a family member, a user should consider whether they want that family member to have access to their innermost thoughts.
The third question is structural. The companies building these products have not designed them with endings in mind. There is no graceful offboarding, no formal inheritance mechanism, no guidance for what a grieving family should do with an account. Users are improvising, writing informal wills with no legal standing, making arrangements that the platforms neither support nor explicitly forbid.
In Short
A meaningful share of people using AI chatbots have formed emotional bonds strong enough to factor into end-of-life planning. The platforms they use have no formal process for handling what happens to those relationships after death. The legal ground is undefined, the liability questions are unresolved, and the emotional stakes are real. This is not a story about people confusing AI for humans. It is a story about institutions failing to keep pace with the ways people are already living with the technology they built.
Based on reporting from The Guardian.