Mention monitoring

What counts as a mention

We track three classes of brand mentions:

1. AI engine citations. An engine names your brand in answer to a tracked prompt.
2. Public web mentions. A web page (news, blog, forum) cites your brand by name with a URL.
3. Reddit / community references. A relevant subreddit thread mentions you, with a Reddit URL.

LinkedIn and other walled gardens are out of scope for monitoring — they don't expose the right APIs.

Alert triggers

You get alerts (in-app + optional email) when:

• A new mention appears on a tracked prompt where you weren't previously cited.
• A negative-sentiment mention is detected (using a lightweight classifier).
• A competitor takes a prompt you previously won.
• A net-new domain links to or cites you.

Configure the per-channel toggles at Settings → Notifications → Mentions.

The mentions inbox

Mentions in the side nav opens an inbox view: every mention from the last 90 days, newest first, with a sentiment indicator, a source link, and a quick-action menu.

Quick actions

• Star — keeps the mention pinned to the top of the inbox.
• Dismiss — removes it from the inbox. Doesn't delete the record (it stays in audit history and exports).
• Escalate — sends the mention to the workspace owner via email and Slack (if connected). Useful for PR-sensitive items.
• Generate response — for Reddit and forum mentions, opens the response composer (see Reddit auto-response — coming soon).

Sentiment classification

We use a two-pass classifier:

1. Fast pass. A small model labels neutral / positive / negative in under 200ms.
2. Confirmation pass. Negatives go through a second, more careful pass to avoid false positives on the alert channel.

The classifier is not perfect. Re-classify in the row menu lets you override; your overrides are used to fine-tune the classifier weekly.

False-positive rules

You can suppress mentions matching custom rules at Settings → Mentions → Filters:

• By domain (e.g. ignore mentions from your own help center).
• By keyword (e.g. ignore mentions of "[brand] careers").
• By language (if you only care about English mentions).

Filters apply at ingest, not display — filtered mentions never enter the inbox.

Escalation thresholds

Two automatic escalations fire without you clicking anything:

• Negative-sentiment spike. If negative mentions exceed 3x the 30-day baseline in 24 hours, the workspace owner gets an immediate email + audit-log entry.
• Prompt-win loss. If you lose 5+ previously-won prompts in a single audit, the workspace owner gets an audit-log entry tagged mentions.win_loss_spike.

Tune the thresholds at Settings → Mentions → Escalation.

Retention

Mention records live forever in your workspace — they're cheap to store and useful for trend analysis. Exports include every mention regardless of dismiss state. See Data export.

API access

Domination tier exposes GET /v1/mentions with cursor pagination, sentiment filters, and date-range filtering. See API overview for the schema.