Message Moderation
Stoked automatically checks messages between prospects and advocates for harmful content before delivering them.
How It Works
Every message a prospect or advocate sends, by text or on the web, is checked by an AI moderation model as it arrives. The check looks for content such as harassment, hate, threats, sexual content, self-harm and violence.
- Not flagged: the message is delivered as normal
- Flagged: the message is held and never delivered
Moderation is always on. There’s no setting to turn it off, and it’s separate from the optional AI summaries and suggested replies.
What Happens to a Flagged Message
- The recipient isn’t texted or emailed about it
- In the advocate site and the community site, both people see a red bubble reading “This message has been flagged.”
- The sender isn’t notified otherwise. Someone texting through Stoked won’t know the message was held.
- Nobody on your team is notified automatically
A flagged message can’t be released. If it was flagged by mistake, ask the sender to rephrase and send it again.
What Admins See
Open the conversation in the admin portal. A flagged message shows:
- A red bubble with the original text, so you can judge it yourself
- A Flagged label where the sent time would be
See Reading a Conversation for the other message statuses.
The reason a message was flagged isn’t shown in the admin portal. If you sync conversations through the API, each message includes its moderation result and the categories that triggered it.
What You Can Do
- Follow up with the sender if the flag looks like a mistake
- Block the prospect from the conversation if the message was abusive. Open the conversation, click the More actions menu and choose Block Prospect. See Blocking and Unblocking Prospects.
- Close the conversation with a fitting close reason. See Closing Conversations.
Related: Conversation Beats · Conversations API