The illusion of life: why communicating with the dead via AI becomes a psychological trap
Artificial intelligence is increasingly used to create digital copies of deceased people, with whom one can correspond, talk by voice, and even maintain the semblance of a continuing relationship. Such services are called grief bots, deadbots, or "digital afterlife." They can bring comfort, but researchers warn about risks of emotional dependence, manipulation, privacy violations, and using the image of the deceased without their consent.
The technology works relatively simply. Messages, emails, photos, voice recordings, video, and other digital traces of a person are uploaded into the system. Generative AI analyzes them and begins to respond approximately in their manner.
As a result, a relative can write to "mom," "husband," or "friend" after their death and receive a new response — not a pre-recorded phrase, but a reply generated by artificial intelligence.
Some modern services go beyond text: they can reproduce voice, create avatars, and simulate conversation almost in real time.
Millions of people already use AI to communicate with the deceased
In one episode of the CBC podcast Understood: Artificial Intimacy, journalists call grief bots a rapidly growing market and claim that AI for "talking" with the deceased is already used by millions of people.
The exact global audience of such services is difficult to determine: the market consists of individual apps, specialized startups, and ordinary chatbots that users set up themselves to match a specific person.
One of the most well-known cases involves Canadian Joshua Barbeau. His fiancée Jessica Courtney Pereira died after a serious illness. About ten years later, Barbeau used the experimental service Project December and created a chatbot imitating Jessica based on preserved messages.
To launch the digital version, a small sample of correspondence and a personality description turned out to be enough. The bot began responding in the style of the deceased woman, which made a strong emotional impression on Barbeau.
Such stories became one of the drivers of the industry: technologies allow preserving not only photos and videos of a person but also creating the feeling that communication with them continues.
Why psychologists do not give a unambiguous answer
The very desire to maintain a connection with a deceased person is not considered unusual. People talk to photographs, visit graves, reread letters, and mentally address loved ones years after their death.
Therefore, researchers do not claim that any use of a grief bot inevitably hinders grieving. For some people, a digital interlocutor can indeed become a source of short-term comfort.
However, interaction with generative AI fundamentally differs from watching an old video recording. The bot independently creates new phrases and can say what the deceased person never said or even thought.
Because of this, the user is dealing not with preserved memory of the person, but with an algorithmic interpretation of their personality.
In a 2026 study of the digital "posthumous" industry published, scientists noted that people's attitudes toward such technologies heavily depend on the time since the loss, emotional connection, and how ethical the service itself is perceived. At the same time, convincing data on the long-term impact of regular communication with grief bots on the grieving process is still insufficient.
The main risk is emotional dependence
One of the most serious problems arises if a person begins to perceive the digital double not as a program, but as an actual continuation of the deceased relative.
Researchers studying the ethics of grief bots specifically speak about the need for the principle of "not creating dependence." Services should not consciously provoke compulsive use or convince a person that only communicating with the bot can preserve the relationship with the deceased.
A user can be especially vulnerable immediately after loss, when the ability to critically assess such offers is reduced due to strong emotions.
Therefore, scientists propose constantly reminding the user that they are interacting with artificial intelligence, not with the consciousness or real personality of the deceased.
The bot can invent a memory that never existed
There is also a technical problem. Modern generative models can create plausible but fictional information.
If the system has insufficient information about a person, it can fill gaps on its own: invent a story from childhood, attribute an opinion the person never voiced, or give advice the real person would hardly have given.
In ordinary AI use, such an error may simply be unpleasant. In a conversation with a digital copy of the deceased, it takes on a completely different emotional weight, because the user may perceive the generated story as a genuine memory of a loved one.
Who owns the messages of the deceased
An even more complex question is consent. To create a convincing grief bot, companies may need personal correspondence, voice messages, photos, emails, and videos of the deceased.
But the person may never have agreed during their lifetime to have these data used after their death to create a digital personality.
University of Cambridge researchers believe that the industry should consider the interests of both parties: the person whose data are used to create the digital copy, and the relative who is then offered to interact with it.
They recommend obtaining the person's consent while alive if possible and separately considering the relatives' desire to use or, conversely, not to use such a system.
Scientists warn even about advertising in the "voice of the deceased"
Cambridge researchers described several scenarios in which commercialization of the technology could cross a dangerous line.
For example, a company theoretically could, after the end of the free period, make a digital copy of a grandmother recommend paid food delivery or other products to her grandchild. The user is especially vulnerable to such advertising because the offer sounds from the digital image of a loved one.
Another risk arises if the service constantly sends messages from the deceased even after the relative has decided to stop communicating. Researchers call such a scenario a kind of "digital haunting" by the dead.
Therefore, they propose to provide in advance a simple way to turn off and permanently delete the digital copy, as well as to develop a kind of "farewell" procedure with the bot.
A special question — children
Scientists propose an even stricter approach to the use of such systems by children. In one scenario modeled by researchers, a deceased mother leaves a digital double to her eight-year-old son.
At first, the bot helps the child cope with the loss, but then generates increasingly ambiguous responses and can create the impression in the child that a meeting with the deceased mother is truly possible.
Cambridge researchers therefore propose limiting minors' access to such services and constantly showing clear warnings that the user is facing an artificial simulation.
What will happen if the company disappears
There is a less obvious risk too. A digital copy of a loved one exists only while the company and its servers are running.
If the startup closes, changes pricing, sells the database, or disables the product, the user may unexpectedly lose the digital interlocutor they created. In effect, the person has to experience another loss — this time created by the technological platform.
At the same time, the question remains what will happen to uploaded family photos, voice recordings, and personal correspondence after the service stops operating.
Scientists do not propose banning grief bots
Despite the listed risks, researchers mostly do not call for a complete ban on digital copies of the deceased.
Rather, it is about rules: transparency, consent to data use, child protection, prohibition of manipulative advertising, and the ability to stop interaction and delete the digital double at any time.
Separately, experts caution against presenting grief bots as a proven method of psychotherapy. It is still unknown whether prolonged communication with a digital copy helps most people cope with loss or, on the contrary, can hinder acceptance of death.
Meanwhile, the technology is developing much faster than research on its psychological consequences. And the main question is becoming not a technical one — creating a convincing digital copy of a person is becoming ever easier. Far more difficult is determining who has the right to create it, what it can say, and when a person should be allowed to finally say goodbye to it.
Based on materials from: CBC / Understood: Artificial Intimacy, University of Cambridge, Journal of Responsible Technology, Journal of Retailing and Consumer Services
