AI chatbots of deceased loved ones may help people cope with grief, study finds
One 32-year-old woman texting a chatbot built from her late grandmother described the experience as visceral
Sixteen people who'd lost someone close to them sat down for Zoom sessions and talked to AI chatbots built from details about their deceased loved ones.
Every single individual said they would do it again. This is according to a new study from the University of Colorado Boulder that is published in the Proceedings of the 2026 Designing Interactive Systems Conference, giving the first in-depth insight into the reality of the experience of something known as "generative ghosts".
The research was led by Jack Manning, a PhD candidate in information science, and Jed Brubaker, an associate professor in the same department. Two different forms of each chatbot were created, one that spoke as the dead in the first person and one that referred to the dead in the third person.
Participants consistently preferred the first-person version, which the researchers call "reincarnation", over "representation".
One 32-year-old woman texting a chatbot built from her late grandmother described the experience as visceral. "I can see her. I can feel her," she told researchers, adding that it gave her the closure she'd been looking for.
Surprisingly, people proved to be quite forgiving regarding errors in the factual domain, but much stricter with errors of language. For example, a chatbot representation of someone's stepfather used the nickname "champ", while the actual stepfather had never used it, which would have almost ruined the experiment right away.
At the same time, users preferred short replies full of emojis rather than long paragraphs, which are typical for AI models, implying that emotional authenticity in these tools is more about rhythm and word choice rather than accuracy.
Companies such as Séance AI and Project December are already offering such services based on diary writings, messages, and social media posts of the deceased. According to Brubaker, the lab was forced to undertake such research, as there was nothing understood about the psychological impact of the technology as the latter had progressed too much.
Manning, who has lost his sister in childhood, said that he still does not trust the uncontrolled use of this technology, especially when it comes to vulnerable people meeting the simulation of their beloved one without any restrictions.
-
US politician slams Sam Altman over 'accept some bad things happening' remark
-
Italy antitrust regulator opens investigation into AI music startup Suno
-
Smart glasses face bans over privacy concerns: Full list of countries
-
Mistral AI new model beats Chinese competitors in cybersecurity, claims CEO
-
UK regulator probes Meta over child safety risks in Instagram's 'Instants' feature
-
OpenAI admits ‘not good enough’ response to Australia government hacks
-
Meta, TikTok and X challenge UK watchdog over safety data: What to know
-
Norway proposes temporary ban on AI glasses in public spaces over privacy fears