Best Practices | AI Verification Help

Overview

Effective use of the AI Verification Service requires understanding not just how the technology works, but also how to integrate it into your decision-making process as an administrator. These best practices have been developed through extensive testing and real-world usage to help you maximize the value of AI-powered copyright verification.

General Best Practices

Always Verify Before Approving

Make AI verification a mandatory step in your registration approval workflow. No registration should proceed to blockchain recording without a verification report.

Set up your workflow to require verification before final approval
Train all administrators on verification importance
Document verification results in approval decisions
Use verification as quality assurance checkpoint

Review Full Reports, Not Just Scores

Overall risk scores provide quick assessment, but detailed findings contain crucial context that may change your decision.

Read executive summary and detailed findings
Check provider-specific analysis for additional insights
Look for patterns across multiple verification types
Consider strength of evidence, not just numerical scores
Pay attention to warnings and red flags

Investigate Medium-Risk Results

Medium-risk scores (31-70) require human judgment. Don't automatically approve or reject these cases without investigation.

Request additional documentation from applicant
Cross-reference with external databases
Consult with other administrators
Look for corroborating evidence
Document your reasoning for final decision

Trust But Verify AI Recommendations

AI recommendations are powerful tools, but not infallible. Use them to inform decisions, not replace human judgment.

Understand why AI reached its conclusion
Look for potential edge cases or unusual situations
Consider factors AI cannot assess (reputation, history)
Override recommendations when justified, but document why
Report false positives/negatives to improve system

Content Submission Best Practices

Ensure Complete Submissions

Verification quality depends on content quality. Incomplete submissions produce less reliable results.

Always Include When Available:

Audio Files: High-quality recordings (320kbps MP3 or better)
Lyrics: Complete lyrical content in text or PDF format
Score Sheets: Musical notation if available
Metadata: Accurate duration, format, and technical details
Album Art: Cover images for additional verification

Validate File Quality

Poor quality files can lead to inaccurate analysis. Verify file integrity before submission.

Check audio files play correctly
Ensure PDFs are readable and not corrupted
Verify images are clear and not pixelated
Confirm file sizes are within limits
Test files on different devices if possible

Use Appropriate File Formats

Supported formats produce better results than converted or unusual formats.

Recommended Formats:

Audio: MP3 (320kbps), WAV (uncompressed), FLAC (lossless)
Documents: PDF for lyrics and scores, TXT for plain text
Images: PNG or JPG for album art and visual content

Decision-Making Framework

Low Risk (0-30): Approve Pathway

Low-risk results indicate legitimate content, but still require basic verification.

Review executive summary for any unusual findings
Verify all required documentation is present
Confirm participant information is accurate
Check for any red flags in detailed findings
If all clear, proceed with approval
Document decision in registration notes

Medium Risk (31-70): Investigation Required

Medium-risk results need deeper analysis before making final decision.

Identify specific concerns from detailed findings
Research external sources (ASCAP, BMI, copyright databases)
Contact applicant for clarification or additional evidence
Consult with senior administrators if needed
Re-run verification if new content provided
Make decision based on totality of evidence
Document investigation process and reasoning

High Risk (71-100): Rejection Pathway

High-risk results typically warrant rejection unless exceptional circumstances exist.

Review specific evidence of infringement or fraud
Verify AI findings are accurate and not false positives
Consider if applicant has legitimate explanation
Request evidence of ownership or authorization
If concerns remain, proceed with rejection
Provide clear explanation to applicant
Document decision with supporting evidence

Special Considerations

AI-generated content (authenticity below 40) should be rejected regardless of other scores
Clear duplicates (probability above 80) should be rejected to prevent double registration
High laundering risk (above 70) suggests deliberate fraud, treat seriously
When in doubt, err on side of caution with rejection or further investigation

Workflow Optimization

Batch Processing Strategy

Efficient batch processing helps manage high volumes without sacrificing quality.

Group similar registration types together
Process low-risk items first for quick wins
Flag medium/high-risk items for detailed review
Schedule dedicated time for complex investigations
Monitor queue to avoid bottlenecks
Use batch approval for low-risk consensus items

Time Management

Allocate appropriate time based on risk levels to balance efficiency and thoroughness.

Recommended Time Allocation:

Low Risk: 5-10 minutes per registration
Medium Risk: 15-30 minutes per registration
High Risk: 30-60 minutes per registration
Investigation: 1-3 hours for complex cases

Documentation Standards

Maintain consistent documentation for audit trails and future reference.

Record verification ID in registration notes
Summarize key findings from report
Document reasoning for approval/rejection
Note any deviations from AI recommendations
Keep communication records with applicants
Track patterns and trends for process improvement

Common Pitfalls to Avoid

Mistakes That Lead to Poor Decisions

Rubber Stamping: Approving all low-risk items without reading reports
Score Fixation: Making decisions based only on overall risk score
Ignoring Context: Not considering applicant history or special circumstances
Provider Bias: Trusting one AI provider over others without justification
Incomplete Documentation: Not recording decision reasoning
Skipping Verification: Bypassing verification for "trusted" submitters

Continuous Improvement

Learn From Experience

Track patterns and outcomes to refine your decision-making process.

Keep records of decisions and their outcomes
Review disputed cases to identify improvement areas
Share insights with other administrators
Identify recurring issues or patterns
Update internal guidelines based on learnings

Provide Feedback

Help improve the AI system by reporting false positives and negatives.

Report incorrect AI assessments to development team
Provide context for edge cases
Suggest improvements to detection methods
Share successful investigation techniques
Contribute to training data quality

Stay Updated

Keep current with service improvements and new detection capabilities.

Review service updates and release notes
Attend training sessions on new features
Test new detection capabilities as they're released
Understand limitations of current technology
Adapt workflow as service evolves

Administrator Checklist

Before Verification

Verify all required files are present and valid
Check file formats and sizes are within limits
Review registration details for completeness
Ensure participant information is accurate

During Review

Read executive summary and overall recommendation
Review all four individual confidence scores
Examine detailed findings for specific evidence
Check provider-specific analysis for consensus
Note any red flags or unusual patterns

Making Decision

Consider all evidence holistically
Apply appropriate decision framework for risk level
Document reasoning for decision
Record verification ID in registration notes
Communicate decision to applicant if rejection

Best Practices for AI Verification