Targeted messages intended to intimidate, demean, or distress another person. Amanda detects direct attacks, repeated negative messages aimed at the same user, and coordinated pile-ons where multiple users target one individual. Context matters here: the same language between friends in a casual match reads differently than an unsolicited message to a stranger.

Sustained negative behavior directed at a specific person over time. Where harassment is often a single incident, bullying is a pattern. Amanda tracks behavioral history across sessions to identify users who repeatedly target the same individuals, even when individual messages appear borderline in isolation.

Explicit or sexually suggestive content that violates platform policies. This includes direct sexual language, unsolicited advances, lewd comments about appearance, and content that sexualizes other users. For children’s platforms, thresholds are set significantly higher and the category integrates directly with grooming detection.

Attempts to deceive, scam, or manipulate other users for financial or personal gain. Amanda detects phishing language, fake giveaway patterns, impersonation attempts, and links to known scam domains. Particularly relevant for marketplaces and platforms where real money or valuable items change hands.

High volume, repetitive, or unsolicited content that degrades the quality of community interaction. Amanda removes up to 95 percent of spam before it reaches your moderation queue, covering bot-generated content, promotional flooding, and coordinated inauthentic behavior.

Content that facilitates or promotes illegal behavior, including drug sales, weapons trafficking, and the sharing of illegal material. Amanda flags language patterns associated with illegal transactions and escalates high confidence cases for immediate review.

Content that puts users at risk of physical harm, including self-harm encouragement, threats of violence, and content that normalizes dangerous behavior. For platforms with younger audiences, this category works alongside mental health alerting to ensure at-risk users are identified early.

Behavior patterns consistent with the grooming of minors, including excessive personal information requests, attempts to move conversation to private channels, and language designed to build inappropriate trust with a young person. This is one of Amanda’s most sensitive detection categories and is built specifically for children’s platforms and family-friendly communities. Cases are escalated immediately for human review.

The sharing of personally identifiable information, including phone numbers, addresses, financial details, and government ID numbers. Amanda detects and flags PII sharing to protect user privacy and support GDPR and COPPA compliance obligations.