> ## Documentation Index
> Fetch the complete documentation index at: https://chainpatrol.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# How We Use AI

> Learn how ChainPatrol's AI-powered detection engine protects your brand from threats in real-time

## Overview

Threats move fast. Phishing sites can be spun up in minutes, targeting your community before traditional security measures can respond. That's why we built an AI-first detection engine that identifies and neutralizes threats within seconds, not hours.

<Info>
  Our approach combines the speed of automation with the precision of human expertise, ensuring your brand stays protected around the clock.
</Info>

## The ChainPatrol Detection Mix

We've engineered the optimal balance between speed and accuracy:

| Detection Type      | How It Works                                                         |
| ------------------- | -------------------------------------------------------------------- |
| **Fully Automated** | Threats that meet all AI confidence thresholds are blocked instantly |
| **Human Review**    | Edge cases are flagged for immediate analyst review                  |

<Note>
  **Why this matters:** Multiple AI and rule-based thresholds must pass before any action is taken. This layered approach prevents false positives while ensuring legitimate threats are caught quickly.
</Note>

## How Our AI Detects Threats

ChainPatrol uses a multi-layered AI approach, combining multiple detection methods for comprehensive threat protection.

### Content Analysis

Our AI analyzes text content across platforms to identify threats:

* **Language Pattern Detection** - Identifies phishing language, scam indicators, and social engineering tactics in posts, messages, and web content
* **Context Understanding** - Recognizes subtle impersonation attempts and fraudulent claims that might fool traditional filters
* **Multi-Platform Monitoring** - Analyzes content across social media, websites, messaging apps, and more

### Visual Detection

Phishing sites often look identical to legitimate ones. Our visual AI catches them:

* **Visual Similarity Analysis** - Compares page appearance against your legitimate sites to catch pixel-perfect impersonation attempts
* **Structural Pattern Matching** - Examines page structure, code patterns, and layouts to identify known phishing templates, even when visuals differ
* **Real-Time Scanning** - Continuously monitors suspicious URLs for visual indicators of threats

### Specialized Detection

Our models are trained specifically for online threats:

* Trained on hundreds of thousands of real threat examples from across the ecosystem
* Continuously updated with new attack patterns as they emerge
* Fine-tuned for specific threats like wallet drainers, fake airdrops, and impersonation scams

## How Detection Works

Here's what happens when a potential threat is identified:

1. **Discovery** - Our engine monitors 50+ platforms 24/7
2. **Analysis** - Multiple AI models analyze the threat independently
3. **Scoring** - Our decision engine aggregates confidence scores
4. **Action** - High-confidence threats are blocked; edge cases go to analysts
5. **Distribution** - Confirmed threats are pushed to all integrated platforms

### The Decision Process

At the core of our system is a sophisticated decision engine that:

* **Multi-Signal Aggregation** - Combines outputs from all detection methods
* **Confidence-Based Scoring** - Weighs evidence to determine threat likelihood
* **Cross-Reference Validation** - Checks against known threat databases and patterns
* **Instant Decisions** - Makes final determinations in seconds

No single signal is enough on its own. Multiple independent checks must pass before a threat is actioned, helping maintain accuracy while preserving speed.

## Platform Coverage

Our AI monitors threats across 50+ platforms, and we can add new integrations within days.

**Core Detection Types:** URLs & Domains, Web Pages, Wallet Addresses, Phone Numbers, Email Addresses, Forms & Surveys

**Social & Communication:** Discord, Twitter/X, Telegram, LinkedIn, YouTube, Reddit, TikTok, Instagram, Threads, Medium, WhatsApp, Facebook, Snapchat, Substack

**App Stores & Extensions:** Google Play, Apple App Store, Chrome Web Store, Firefox Add-ons, ChatGPT Plugins, Microsoft Store, Opera Extensions

**Blockchain & Social Platforms:** OpenSea, ENS, IPFS, Farcaster, Blur, Magic Eden, DeBank, Bluesky, Primal, DeSo

**Need coverage for a platform we don't list?** Contact us. New integrations can be added within days.

## Real-Time Threat Distribution

When our AI detects a threat, it doesn't just sit in a database. We push it directly to the platforms your users rely on:

### Browser Protection

**Google Safe Browsing** - Our integration means threats are blocked in Chrome, Edge, and Safari, reaching billions of users worldwide.

### Wallet Protection

Direct integrations with leading wallets provide real-time warnings at the point of transaction:

* **MetaMask** - Core integration partner protecting millions of users
* **Phantom** - Protecting the Solana ecosystem
* **WalletConnect** - Real-time threat prevention
* **20+ More Wallets** - Comprehensive ecosystem coverage

**The result:** Threats detected by ChainPatrol are blocked across the entire ecosystem within seconds.

## Continuous Learning & Improvement

Our AI gets smarter every day through:

**Human-in-the-Loop Training** - Every analyst decision feeds back into our AI. When our team reviews edge cases, that knowledge improves future detection accuracy.

**Threat Intelligence Networks** - We're active contributors to major threat intelligence communities, giving us early access to emerging threats and allowing us to share our detections with the broader security ecosystem.

**Expanding Datasets** - Our threat database grows daily with new examples, ensuring our AI stays ahead of evolving attack techniques.

## Key Takeaways

* Speed without sacrifice: Automated detection handles high-confidence threats instantly while human review prevents false positives on edge cases
* Layered protection is stronger: Multiple AI signals must agree before action is taken, which prevents both missed threats and wrongly blocked assets
* Training on specialized threats matters: Generic phishing detection misses wallet drainers and fake airdrop patterns that our specialized models catch
* Direct integrations multiply impact: A single detection can protect users across browsers, wallets, and platforms within seconds of confirmation