How a griefbot creates replies
The system combines a language model with selected instructions and memories. It may retrieve a relevant story or phrase and then generate a new response, which means a familiar-sounding answer can still include an unsupported inference.
Why source labels matter
Users should be able to distinguish a direct quotation, a family-provided memory, a profile instruction, and AI-generated interpretation. Without provenance, new text can be mistaken for an authentic statement.
Consent extends beyond the deceased person
Messages and stories often include living family members, friends, or colleagues. Their privacy and rights still matter, and access should be limited to people covered by the intended memorial use.
Design for reflection and an easy exit
The experience should keep the AI label visible, avoid dependency-promoting language, provide pause and deletion controls, and direct distressed users toward human support rather than presenting itself as therapy.
Explore ExChat memorial AI. Continue in ExChat's mobile-first creation flow.
Continue ↗