Visitor Memory

Visitor Memory#

Visitor Memory lets your chatbot remember useful facts a returning visitor has shared — their preferences, language, or an open question they had last time. When the same visitor starts a new conversation, the assistant already has that context. It does not need to ask again.

Memory is available on all plans. It is off by default and must be enabled per widget.

How it works#

At the end of each conversation, the assistant extracts a small set of distilled facts — things like a stated preference or an open question the visitor left with. These are stored and loaded silently at the start of the visitor's next conversation.

Memory stores preferences and open threads. It does not store raw conversation transcripts. The assistant uses memory naturally without announcing it.

Facts are capped at a small number per visitor to keep the feature lightweight. When the cap is reached, older facts are replaced by newer ones.

Enabling visitor memory#

Open the widget editor in the Console and go to Visitor Context. Toggle Visitor memory on and publish the widget.

The toggle is per widget. An account with multiple widgets can enable memory on some and leave it off on others. The widget's privacy notice is the consent surface — review it before enabling.

Retention#

Each fact has a fixed 90-day expiry. The timer resets when a returning visitor refreshes the fact. A visitor who does not return within 90 days will have their facts expire automatically — no manual cleanup needed.

Privacy and data handling#

Memory is scoped to the widget it was captured under. It never crosses accounts or widgets. Email addresses and other identifiers are filtered before storage — only preference and context facts are kept.

Disabling visitor memory on a widget stops new facts from being captured. Existing facts are still subject to their 90-day TTL.