STOP and the do-not-contact list

The do-not-contact list protects people who have opted out and prevents proactive outreach within the workspace. A STOP sent by the customer carries more weight than an ordinary exclusion entry: it records withdrawal of consent to receive messages.

#How is STOP detected?

The current detector version is tg-stop-v1. It checks the entire message, not a STOP word somewhere in a long sentence.

Before matching, the system normalizes Unicode, converts text to lowercase, replaces punctuation and emoji with spaces, and collapses whitespace. Vietnamese diacritics are preserved. The spelling “huỷ” is specifically converted to “hủy”.

A STOP phrase may have up to 2 polite words after it, such as “ạ”, “nhé”, or “please”. Content longer than 60 characters after normalization is not recognized as STOP.

Some unaccented spellings are declared separately, such as khong quan tam and huy dang ky. This does not mean the detector removes diacritics from every sentence.

#Which messages are checked?

STOP runs before your automation rules and checks only live incoming direct private messages:

• It does not check messages sent by your accounts.
• It does not check historical messages retrieved during sync.
• It does not check system notices or forwarded messages.
• It does not use group messages to conclude that someone has sent STOP.
• It does not apply to people identified as Internal.

If a batch contains several matching STOP messages, the system selects the newest one.

#What does STOP block?

STOP creates or updates a do-not-contact entry with the source opt_out. If no newer message from the customer has arrived at the account being checked, the gate blocks both outreach and replies. Calling a message a “reply” does not override withdrawn consent.

If the customer initiates contact again after STOP, the gate may allow replies based on the incoming message time for that account. The do-not-contact entry does not automatically disappear. Proactive outreach remains blocked.

For an ordinary do-not-contact entry, replying to someone who has previously messaged may be allowed. Outreach actions such as inviting someone to a group must still check the list.

The list applies across the workspace. Switching to another sending account is not a way to continue approaching an excluded person.

#What evidence is needed to remove someone?

For entries created by STOP, the system requires written evidence: at least 5 characters, up to 300 characters. An AI connection needs human confirmation to remove the entry.

Meeting the length requirement does not mean the evidence has been verified as valid consent. You remain responsible for clearly recording the basis, such as the customer explicitly asking to continue the conversation.

An entry with the data-erasure request source (erasure) cannot be removed through this mechanism.

#Related

• Internal accounts
• Manage people
• Safe sending for agents