Choosing the right combination of prompts and AI models is crucial for effective monitoring. This guide provides strategic approaches to optimize your analysis configuration.
Understanding the Trade-offs
Every analysis involves balancing three factors:
- Coverage - How comprehensive your monitoring is
- Cost - Quota usage and execution resources
- Time - How long execution takes
More prompts and models = better coverage but higher cost and longer execution.
Prompt Selection Strategies
Comprehensive Baseline Approach
Best for: Initial monitoring, establishing benchmarks
- Include all automatic prompts
- Add 5-10 custom prompts
- Cover all categories evenly
- Typical range: 30-50 prompts
- Execute monthly or quarterly
Pros
- Complete picture of AI presence
- Identifies unexpected strengths/weaknesses
- Establishes comprehensive baseline
Cons
- Higher quota usage
- Longer execution time
- More data to analyze
Focused Category Approach
Best for: Category-specific optimization, targeted monitoring
- Select all prompts from one category
- Add related custom prompts
- Typical range: 10-20 prompts
- Execute weekly or bi-weekly
Example: Brand-Focused Analysis
- All Brand category automatic prompts
- Custom competitive comparison prompts
- Brand awareness and recognition prompts
- 15-20 prompts total
Pros
- Deep insights in specific area
- Lower cost per execution
- Faster execution
- Easier to analyze results
Cons
- Misses broader picture
- May overlook related issues in other categories
Priority-Based Approach
Best for: Quota-constrained plans, efficient monitoring
- Identify 10-15 most critical prompts
- Include mix of all categories
- Focus on highest-value questions
- Execute frequently (weekly)
How to Identify Priority Prompts
- Prompts related to key business metrics
- Questions your customers ask most
- Competitive positioning prompts
- Prompts showing biggest variation
Pros
- Efficient quota usage
- Quick execution
- Focused insights
Cons
- May miss emerging issues
- Less comprehensive coverage
Campaign-Specific Approach
Best for: Marketing campaigns, product launches, specific initiatives
- Custom prompts tailored to campaign
- Related automatic prompts
- Competitive prompts for context
- Typical range: 8-15 prompts
- Execute before, during, and after campaign
Example: Product Launch Campaign
- Custom prompts about new product
- Product category automatic prompts
- Competitive product comparison prompts
- Brand awareness prompts for context
For campaigns, create a dedicated analysis that runs throughout the campaign lifecycle for consistent tracking.
AI Model Selection Strategies
Available AI Models
Microscope.ai typically supports:
- GPT models (OpenAI) - ChatGPT
- Gemini models (Google) - Gemini/Bard
- Claude models (Anthropic) - Claude
- Perplexity AI - Perplexity search
All Models Approach
Best for: Comprehensive monitoring, initial baseline
- Query all available AI models
- Most complete picture
- Understand model-by-model differences
- Identify which models represent you best/worst
Pros
- Complete cross-model insights
- Identifies model-specific issues
- Supports comprehensive reporting
Cons
- Highest cost
- Longest execution time
- Most data to analyze
Recommended For
- First analysis of a project
- Monthly or quarterly comprehensive reviews
- Executive reporting
- Optimization impact assessment
Primary Models Approach
Best for: Regular monitoring, efficient execution
- Select 2-3 most important models
- Focus on models your audience uses
- Typical: GPT + Gemini
- Execute weekly
How to Choose Primary Models
- Market share in your target audience
- Which models show most variation
- Client or stakeholder preferences
- Models with best baseline performance
Pros
- Balanced cost and coverage
- Faster execution
- Sufficient for trend tracking
Cons
- May miss model-specific changes
- Less comprehensive than all-models
Single Model Approach
Best for: Deep-dive investigations, model-specific optimization
- Query only one AI model
- Useful for troubleshooting
- Focus optimization efforts
- Execute frequently
When to Use
- Investigating model-specific issues
- Testing optimization impact on one model
- Model-focused A/B testing
- Very tight quota constraints
Pros
- Lowest cost
- Fastest execution
- Highly focused insights
Cons
- Incomplete picture
- May not represent overall AI landscape
Rotating Models Approach
Best for: Long-term monitoring with quota constraints
- Rotate which models you query
- Week 1: GPT + Gemini
- Week 2: Claude + Perplexity
- Provides breadth over time
Pros
- Balances cost and coverage
- Ensures all models monitored eventually
- Works within quota limits
Cons
- Comparisons less direct
- Harder to track trends per model
- Requires more planning
Combining Strategies
Recommended Monitoring Mix
For most businesses, a balanced approach works best:
Monthly Comprehensive
- All automatic prompts + key custom prompts
- All AI models
- Comprehensive baseline tracking
Weekly Focused
- 15-20 priority prompts
- 2-3 primary models
- Track most important metrics
Ad-Hoc Targeted
- Category or campaign-specific prompts
- Relevant models
- As needed for specific questions
Selection Best Practices
Prompt Selection
- Start broad, then focus - Begin with comprehensive, narrow based on findings
- Balance categories - Don't focus only on one category
- Include custom prompts - Add 20-30% custom to automatic
- Remove underperformers - Drop prompts that yield no insights
- Test variations - Try different phrasings of key questions
Model Selection
- Know your audience - Focus on models they use
- Establish baseline with all - First analysis should include all models
- Prioritize based on data - Focus on models showing most variation
- Reassess quarterly - Model importance can shift
- Consider market trends - Monitor emerging popular models
Common Selection Scenarios
New Project - First Analysis
- Prompts: All automatic + 5 key custom prompts
- Models: All available models
- Purpose: Establish comprehensive baseline
Ongoing Brand Monitoring
- Prompts: All Brand category + Trust category + select Product prompts
- Models: GPT + Gemini (primary audience models)
- Frequency: Weekly
Product Launch
- Prompts: Product-specific custom prompts + Product category automatics
- Models: All models
- Frequency: Daily during launch, weekly after
Competitive Analysis
- Prompts: Comparison and positioning prompts
- Models: Models where competition performs well
- Frequency: Monthly or quarterly
Quota-Constrained Monitoring
- Prompts: 10 highest-priority prompts only
- Models: Top 2 models by audience use
- Frequency: Weekly or bi-weekly
Optimizing Based on Results
After First Execution
- Review which prompts provided valuable insights
- Identify which models showed most variation
- Note any redundant prompts
- Find gaps in coverage
Refinement Actions
- Remove prompts with consistently generic answers
- Add custom prompts for identified gaps
- Focus models based on performance differences
- Adjust balance between categories
Continuous Improvement
- Reassess selection quarterly
- Track which prompts drive action
- Monitor cost vs. value
- Test new prompts regularly
Next Steps
With strategic selection skills, you're ready to:
- Run optimized analyses
- Set up efficient recurring monitoring
- Maximize insights within quota limits
- Interpret and act on analysis results