An AI-generated response is a simulation. It does not establish what a person would have said, preserve their consciousness, or create contact with someone who has died.
Distinguish an archive from a simulation
An archive preserves actual letters, recordings, and photographs. A generative system creates new material from patterns and instructions. Even when a response sounds familiar, it can invent memories, opinions, and events. Keep original material clearly separated from generated text or audio.
Ask whose information is involved
A family collection may contain private details about living people as well as the person being remembered. Before uploading anything, consider permission, the deceased person’s expressed wishes, the service’s terms, and what other people reasonably expect to remain private. Do not assume ownership of a copy means every use is appropriate.
The Canadian privacy regulators’ AI principles are a useful reference for thinking about personal-information handling. They are not a determination of the legal rights in a particular family archive.
Label generated material plainly
If a memorial includes generated text or audio, make that fact obvious. Do not present invented statements as quotations or use a likeness to imply consent, endorsement, or a new personal message. Decide who can access the material and how it can be removed.
Choose an approach that feels right for the people involved
A simple, permission-based collection of authentic memories may meet the need without a conversational simulation. People can respond differently to memorial technology; no emotional benefit is guaranteed. If using it is distressing or interfering with daily life, stepping back and seeking support from trusted people or a qualified professional may help.
Sources & review
Primary sources checked on . The checklists and planning examples are AI-assisted editorial guidance, not source quotations or reported client results.
This article was revised on October 3, 2026. The update replaces unverified anecdotes and broad claims with practical guidance and primary-source references. The original article URL is preserved.
Originally published: .