How Yappit's Moderation Actually Works

Random video chat has a reputation problem, and it earned it. Omegle, the platform that popularized talking to strangers online, shut down permanently in November 2023 after years of lawsuits tied to child safety failures and an inability to moderate at scale. Ome.tv continues to operate but faces ongoing criticism about the frequency of explicit content users encounter. In both cases the core issue was the same: the platforms either couldn't or didn't invest in moderation systems that worked in real time.

Yappit exists because we think the concept — genuine connection with a stranger — is worth saving. But it only works if the environment is actually safe. This page explains, in plain terms, exactly how our moderation system works, what it catches, and what it cannot.

Text Moderation: Two Layers, Not One

Every text message sent on Yappit's Chat Yap mode passes through two separate filters before it ever reaches the other person. This is not a single blocklist. It is two distinct systems that catch different kinds of violations.

Layer 1: Instant client-side keyword detection. This runs inside your browser the moment you hit send. It checks your message against a maintained list of explicit terms, slurs, and known harmful phrases. It doesn't just match exact words — it normalizes for common evasion tactics like leetspeak substitution (replacing letters with numbers), homoglyph characters, repeated characters, and space-separated spelling. If you type "s.e.x.t" or "s3xt," it catches that the same way it catches the unmodified word.

Layer 2: AI-powered context analysis. Some messages are harmful without containing any single blocked word. "Come meet me alone" is technically clean text, but the intent can be predatory. The second layer uses AI to evaluate the context and meaning of a message, not just its individual words. This catches manipulation, grooming language, and threatening messages that a keyword list would miss entirely.

Messages that trigger either layer are blocked before delivery. The sender receives a warning. What gets blocked: explicit sexual content, threats of violence, hate speech, requests for personal contact information from minors, and grooming patterns. What does not get blocked: disagreements, mild profanity in casual context, unpopular opinions. We moderate for safety, not for tone.

Video Moderation: AI That Runs in Your Browser

This is where Yappit does something most platforms don't. Our video moderation uses an AI model called NSFWJS that runs entirely inside your browser. Here's what that means in practice.

The AI samples a frame from the video feed roughly every 11 seconds. It analyzes that frame locally — on your device, in your browser — and classifies it across several categories. If the combined probability of explicit content exceeds our threshold, the system acts immediately.

When the system detects a violation on your own camera, it warns you first. A second violation triggers a stronger warning and ends the session. A third triggers a temporary hold on your account. When a violation is detected on the other person's camera, the call is ended immediately to protect you.

This approach was a deliberate architectural choice. Server-side video scanning would mean routing all video through our infrastructure, which would mean we could see your video. We didn't want that. Client-side AI means the moderation works, and your privacy stays intact. If you want a random chat without nudity, the architecture is what delivers it — not just a policy.

The Strike System

Violations follow a clear escalation path. This is not arbitrary — every user gets the same treatment regardless of their subscription tier.

First violation: You see a warning message. Your current session may be ended depending on severity. No lasting consequence on your account yet. This exists because people sometimes trigger filters accidentally — an ambiguous message, a momentary camera issue. We'd rather warn than punish a first-time edge case.

Second violation: A temporary suspension. You'll be unable to start new matches for a set period. This is the system telling you the first warning was not a suggestion.

Third violation: A permanent ban from the platform. Your Google account is flagged and cannot be used to access Yappit again. Because we use Google OAuth for verified sign-ins, you cannot simply create a new throwaway account to get around this. That's the point.

The strike system applies across all three modes — Video Yap, Chat Yap, and Voice Yap. A violation in text chat counts the same as a violation on video.

What We Cannot Catch

No moderation system is perfect, and we're not going to pretend ours is. Here is what our current system does not reliably detect:

• Subtle manipulation: A conversation that slowly becomes inappropriate over many messages may not trigger either text filter if each individual message is technically clean.
• Off-platform contact: If someone verbally tells you a phone number or social media handle during a video call, our text filters cannot catch that because it was never typed.
• Context-dependent harm: Sarcasm, coded language within specific communities, and inside references that carry threatening meaning only to specific people are extremely difficult for any automated system.
• Audio content: Our current moderation covers text and video frames. Audio-only content in Voice Yap is not currently scanned by AI. We rely on user reports for voice-only violations.

We are honest about these gaps because pretending they don't exist would make you less safe, not more. Knowing what the system cannot catch helps you make better decisions about your own safety.

Reporting What the System Missed

Every page on Yappit has a report function. If you experience something that the automated system didn't catch — or if you simply feel uncomfortable and want to flag a conversation — use it. Reports are reviewed and result in the same strike system described above.

You can also visit our report a violation page directly if you want to file a report after a session has ended.

Moderation is not a feature we shipped and forgot. It's an ongoing commitment. We update our keyword lists, retrain our detection thresholds, and review reports continuously. The goal is not a random chat without nudity by accident — it's a moderated anonymous chat platform by design. If you want to understand what behavior we expect from everyone on Yappit, read our community guidelines. If you want to know exactly what data we do and don't collect while all of this runs, read our privacy explanation.