How MerchantGuard's AI Works
Complete transparency about our AI technology, training data, methodology, and limitations
Last Updated: October 24, 2025
Overview
MerchantGuard uses artificial intelligence and machine learning to provide educational risk assessments and personalized guidance for payment processing compliance. This page explains our methodology, training data, and limitations.
Our AI Technology
AgentGuard™ AI System
MerchantGuard is powered by AgentGuard™, our proprietary AI system that combines:
Large Language Models (LLMs)
For natural language understanding and conversational guidance
Machine Learning Risk Models
For GuardScore calculations and risk predictions
Pattern Recognition Systems
For compliance monitoring and alert generation
Knowledge Retrieval
For accessing payment processing policies and best practices
Training Data
Our ML models are trained on:
- 70,000+ synthetic merchant scenarios: Simulated merchant profiles, transaction patterns, and risk outcomes
- Agent conversation patterns: Thousands of synthetic customer service and compliance advisory interactions
- Public payment processor policies: Publicly available terms of service, compliance guides, and regulatory documentation
- Industry research: Academic papers, compliance frameworks, and payment industry best practices
Important: Our models are trained on synthetic and publicly available data. We do NOT train on: Real merchant transaction data (unless explicitly provided by you for your own assessment), Proprietary processor underwriting criteria, or Confidential merchant information from other users.
How GuardScore Works
When you submit payment data or answer questions, our system:
Analyzes your inputs
Processing volume, chargeback rate, industry, region, entity structure, etc.
Calculates risk metrics
Compares your metrics against known thresholds (VAMP, regional limits, industry benchmarks)
Generates risk score
ML model predicts risk level (0-100 scale) based on training data patterns
Matches processors
Compares your profile against processor acceptance criteria database
Provides recommendations
Generates personalized action plan based on your specific situation
PSP Matching Methodology
Our PSP matching system considers:
- Entity structure compatibility: Which processors accept your legal entity type (LLC, S.A. de C.V., etc.)
- Regional requirements: Processors licensed to operate in your processing regions
- Industry acceptance: Processors that accept your specific industry (CBD, gaming, crypto, etc.)
- Risk tolerance: Processors known to work with businesses at your risk level
- Published rates: Publicly available pricing information
PSP match results are estimates based on publicly available information and ML predictions. Actual approval decisions are made independently by payment processors.
Compliance Alert System
Our AI monitors 45+ sources including:
- Payment processor documentation (Stripe, PayPal, Square, etc.)
- Card network updates (Visa, Mastercard, Amex)
- Regulatory announcements (CFPB, FTC, state regulators)
- Industry forums and communities
- Payment industry news sources
When relevant changes are detected, our system:
- Analyzes the impact on different merchant profiles
- Identifies which merchants are affected
- Generates personalized alerts explaining the impact on YOUR specific business
- Provides recommended actions based on your profile
Limitations and Important Disclaimers
✓ What Our AI Can Do
- Provide educational risk assessments
- Calculate risk scores based on known compliance thresholds
- Identify potential compliance issues
- Suggest mitigation strategies
- Match you with potentially compatible processors
- Monitor public compliance sources for changes
- Explain payment processing concepts
✗ What Our AI Cannot Do
- Guarantee processor approval or account retention
- Access processor underwriting systems
- Predict exact processor decisions
- Provide legal or financial advice
- Replace professional compliance consultation
- Guarantee prevention of account freezes or terminations
- Override processor terms of service
Accuracy and Reliability
Our AI predictions are educational estimates, not guarantees.
- GuardScore accuracy: Our risk scores are based on publicly known thresholds and ML predictions. Actual risk depends on many factors we cannot access.
- PSP matching accuracy: We use publicly available information and ML estimates. Actual processor acceptance criteria may differ.
- Alert timeliness: We monitor sources 24/7, but cannot guarantee we will catch every policy change instantly.
- Recommendation effectiveness: Our suggestions are based on industry best practices, but results vary by situation.
Final outcomes depend on:
- Processor underwriting policies (which are proprietary)
- Your specific implementation of recommendations
- Factors outside our visibility (account history, processor relationships, etc.)
- Changes in processor policies after our recommendations
- Accuracy of information you provide
Data Processing
How We Use Your Data
When you use MerchantGuard:
- Assessment Data: Information you provide (volume, chargeback rate, industry, etc.) is processed by our AI to generate your personalized GuardScore and recommendations.
- Conversation Data: Your questions and interactions with Guard are processed to provide personalized responses.
- CSV/File Uploads: If you upload payment processor statements, we analyze them to extract relevant metrics (transaction volume, chargeback counts, etc.).
Your data is used ONLY to provide services to you. We do NOT:
- Train our models on your specific data without permission
- Share your data with processors or third parties
- Sell your data
- Use your data to train models that benefit other users
See our Privacy Policy for complete details.
AI Model Improvements
We improve our AI through:
- Aggregate anonymized metrics: General patterns (e.g., "merchants in X industry have Y% approval rates")
- User feedback: When you rate responses or report issues
- Public data updates: New processor policies, regulatory changes, etc.
- Synthetic scenario generation: Creating new training examples based on industry patterns
We do NOT use your specific merchant data to train our models unless you explicitly opt-in to a data contribution program (not currently offered).
Independent Assessment
MerchantGuard is an independent educational platform.
We are NOT:
- Affiliated with any payment processor
- Commissioned on processor referrals
- Receiving compensation for processor recommendations
- Acting as a processor agent or ISO
Our recommendations are based on:
- Publicly available information
- ML predictions from our training data
- Industry best practices
- Your specific profile inputs
Processor matches are educational suggestions, not endorsements or guarantees.
Questions?
For questions about our AI methodology:
- Email: ai@merchantguard.ai
- Technical Documentation: docs.merchantguard.ai
- Support: support@merchantguard.ai
Updates to This Page
We may update this methodology page as our AI systems evolve. Last updated: October 24, 2025. Material changes will be noted in our changelog and users will be notified.
MerchantGuard™