The 5 Research Modes
Not all research is the same. Here are the 5 distinct modes and when to use each.
Mode 1: Exploratory Research
Use When: You're new to a topic and need a comprehensive overview
Example Prompt:
I'm completely new to quantum computing. Conduct deep research
to give me:
1. A foundational understanding suitable for someone with
basic computer science knowledge
2. The current state of the field (2024-2025)
3. Key players (companies, researchers, institutions)
4. Major milestones achieved and upcoming
5. Practical applications emerging now
6. What I should learn first if I want to enter this field
Structure this as a beginner's guide that I can reference
over the next month as I learn.
Why It Works:
- Clearly states your knowledge level
- Requests specific structural elements
- Sets a practical use case (reference guide)
Mode 2: Competitive Analysis Research
Use When: You need to understand a market, competitors, or options
Example Prompt:
Conduct a comprehensive competitive analysis of the top 5
project management tools for software development teams
(10-50 people):
For each tool, research and provide:
- Core differentiators
- Pricing model breakdown
- Integration ecosystem
- User satisfaction (verified reviews from G2, Capterra)
- Recent feature releases (last 6 months)
- Limitations and common complaints
- Best suited for (team type/workflow)
Present as a comparison matrix followed by detailed analysis.
Include your confidence level for each data point
(high/medium/low) based on source quality.
The "Banana" Element:
"Include your confidence level" – This forces Gemini to evaluate its own research quality.
Mode 3: Technical Deep Dive Research
Use When: You need detailed technical understanding
Example Prompt:
I'm implementing a real-time collaborative editing feature
(like Google Docs). Conduct deep technical research on:
1. CRDT vs OT algorithms
- How each works technically
- Performance characteristics
- Implementation complexity
- Real-world usage examples
2. Current state-of-the-art approaches
- What are companies like Figma, Notion using?
- Any open-source implementations worth studying?
3. Specific technical challenges
- Handling offline/reconnection
- Conflict resolution edge cases
- Performance at scale (1000+ concurrent users)
4. Recommended architecture
- Based on your research, what would you recommend?
- Include trade-offs of the recommendation
Target audience: Senior software engineer
Include code examples or pseudocode where helpful
Key Elements:
- Specific technical context
- Comparative analysis requested
- Real-world examples required
- Clear target expertise level
- Actionable output (recommendation)
Mode 4: Fact-Finding Research
Use When: You need verified facts, statistics, or data
Example Prompt:
I need verified statistics and facts about electric vehicle
adoption for a business presentation. Research and provide:
REQUIRED DATA (with sources):
1. Global EV sales 2023 vs 2024 (year-over-year growth)
2. Top 5 EV markets by volume
3. Average EV price trends (2020-2024)
4. Charging infrastructure growth rates
5. Battery cost per kWh trends
6. Consumer adoption barriers (survey data)
REQUIREMENTS:
- Only use data from: IEA, BloombergNEF, official government
statistics, peer-reviewed sources
- Provide the specific report/document name for each statistic
- Note the data collection date
- Flag any statistics that have conflicting sources
Format as a data reference sheet I can cite in my presentation.
Why This Works:
- Specifies acceptable sources
- Requires citation details
- Asks for conflict flagging
- Clear output format
Mode 5: Decision Support Research
Use When: You need to make an informed decision
Example Prompt:
I'm deciding whether to migrate our company's infrastructure
from AWS to Google Cloud. We currently spend ~$50K/month on AWS.
Conduct research to help me make this decision:
CURRENT STATE ANALYSIS:
- What are the typical pain points companies face with AWS at
this scale?
- What triggers usually cause companies to migrate?
MIGRATION ANALYSIS:
- Documented case studies of similar migrations
- Typical migration timelines and costs
- Common pitfalls and how to avoid them
COST COMPARISON:
- GCP vs AWS pricing for typical workloads
- Hidden costs to consider
- Potential savings (realistic ranges from case studies)
RISK ASSESSMENT:
- What could go wrong?
- Vendor lock-in considerations
- Service parity issues
RECOMMENDATION FRAMEWORK:
- What questions should I be asking?
- What factors should weigh most heavily?
- Red flags that would suggest NOT migrating?
I need to present a recommendation to leadership with confidence.
The Secret Prompting Framework
After testing hundreds of prompts, I developed the PRECISE Framework for Deep Research.
The PRECISE Framework
P - Purpose : Why do you need this research?
R - Role : Who is the target audience?
E - Explicit : What specific questions need answering?
C - Constraints : What limitations or requirements exist?
I - Information : What sources or types of data are needed?
S - Structure : How should the output be organized?
E - Evaluation : How will you verify the quality?
Framework in Action
Bad Prompt:
"Research AI trends for 2025"
PRECISE Prompt:
PURPOSE:
I'm writing a trend report for our company's annual strategy
meeting. Leadership will use this to prioritize AI investments.
ROLE:
Target audience is C-suite executives with business backgrounds
but limited technical depth. They need strategic insights,
not technical details.
EXPLICIT QUESTIONS:
1. What AI capabilities are becoming production-ready in 2025?
2. Which industries will see the biggest AI impact this year?
3. What are the realistic ROI expectations for AI implementations?
4. What infrastructure investments should companies prioritize?
5. What are the regulatory considerations to watch?
CONSTRAINTS:
- Focus on enterprise AI, not consumer applications
- Emphasize practical, implementable trends
- Avoid hype; stick to evidence-based predictions
INFORMATION SOURCES:
- Gartner, Forrester, McKinsey reports
- Earnings calls from major AI companies
- Published case studies from 2024
- Academic research reaching commercial application
STRUCTURE:
- Executive summary (1 page)
- Top 5 trends with business impact analysis
- Industry-specific breakdown (Finance, Healthcare, Manufacturing)
- Investment recommendation matrix (High/Medium/Low priority)
- Risk factors to monitor
EVALUATION:
- Include confidence level for each prediction
- Note where experts disagree
- Provide counter-arguments to main conclusions
Advanced Techniques
Technique 1: The Iteration Loop
Don't stop at one query. Use this loop:
Round 1: Broad research
↓
Review results, identify gaps
↓
Round 2: "Based on your research, I notice [X] wasn't covered.
Deep dive specifically into [X]."
↓
Review, identify contradictions
↓
Round 3: "Your research showed [A] but also [B]. These seem
contradictory. Research this specific conflict and
help me understand the nuance."
↓
Round 4: "Now synthesize all rounds into a final comprehensive
analysis."
Technique 2: The Devil's Advocate
You've just provided a comprehensive analysis recommending [X].
Now, conduct fresh research specifically looking for:
1. Cases where [X] failed
2. Critics of [X] and their strongest arguments
3. Alternative approaches that succeeded where [X] didn't
4. Hidden risks or downsides not commonly discussed
Be thorough—I need to anticipate counter-arguments.
Technique 3: The Source Triangulation
For the key claim that "[specific claim]," verify this by:
1. Finding the PRIMARY source (original research/data)
2. Finding 2-3 SECONDARY sources that cite it
3. Finding any sources that CONTRADICT it
4. Evaluating the credibility of each source
Present your confidence assessment with reasoning.
Technique 4: The Time-Bound Research
Research how [topic] has evolved over time:
- 2020: What was the state/consensus?
- 2022: What changed?
- 2024: Current state
- 2025 predictions: Where is it heading?
For each time period, cite sources from that era to show
the actual contemporary thinking, not retrospective analysis.
Technique 5: The Expert Perspective Gathering
Research [topic] from multiple expert perspectives:
1. ACADEMIC PERSPECTIVE:
What does peer-reviewed research say?
2. PRACTITIONER PERSPECTIVE:
What do people actually implementing this say?
3. CRITIC PERSPECTIVE:
What are the strongest criticisms?
4. INDUSTRY ANALYST PERSPECTIVE:
What do Gartner/Forrester/etc. predict?
5. END-USER PERSPECTIVE:
What do actual users report (forums, reviews, case studies)?
Note where these perspectives align and where they diverge.
Real-World Case Studies
Case Study 1: Market Entry Research
Scenario: A startup wanted to enter the AI writing assistant market.
Research Prompt Used:
Conduct comprehensive market research for entering the
AI writing assistant market in 2025:
MARKET ANALYSIS:
- Total addressable market (TAM) size
- Current growth rate
- Key market segments (B2B, B2C, specific verticals)
COMPETITIVE LANDSCAPE:
- Top 10 players by market share
- Their positioning and differentiation
- Pricing strategies across the market
- Recent funding rounds and valuations
CUSTOMER INSIGHTS:
- Primary use cases driving adoption
- Key pain points with existing solutions
- Willingness to pay research
- Switching costs and barriers
OPPORTUNITY ANALYSIS:
- Underserved segments
- Emerging use cases
- Geographic opportunities
- Partnership possibilities
ENTRY BARRIERS:
- Technical requirements
- Capital requirements
- Regulatory considerations
- Network effects to overcome
Provide specific, cited data points wherever possible.
Include your assessment of the opportunity attractiveness.
Result: A 15-page market analysis that would have taken a consultant team weeks to produce.
Case Study 2: Technical Architecture Decision
Scenario: Engineering team choosing between microservices and modular monolith.
Research Prompt Used:
Our team (15 developers) is building a B2B SaaS platform
for inventory management. We're deciding between microservices
and modular monolith architecture.
Research this decision comprehensively:
TECHNICAL COMPARISON:
- Performance characteristics of each
- Scalability ceilings and when they matter
- Development velocity impact
- Testing complexity
- Debugging and observability
REAL-WORLD EVIDENCE:
- Companies that chose microservices at our scale
(outcomes, lessons learned)
- Companies that chose monolith and scaled successfully
- Companies that migrated from one to the other (why, how)
TEAM CONSIDERATIONS:
- What team size/structure favors each approach?
- Hiring implications
- Onboarding new developers
COST ANALYSIS:
- Infrastructure costs at our scale
- Operational complexity costs
- Development speed impact on time-to-market
EXPERT OPINIONS:
- What do respected engineers (Martin Fowler, etc.) recommend?
- Are there recent shifts in industry thinking?
DECISION FRAMEWORK:
- What factors should we weigh most heavily?
- What would you recommend for our specific situation?
Case Study 3: Investment Due Diligence
Scenario: Angel investor evaluating a startup.
Research Prompt Used:
I'm conducting due diligence on a Series A investment in
[Company Name], a fintech startup in the embedded finance space.
COMPANY RESEARCH:
- Founding team backgrounds and track records
- Previous funding and investors
- Product offering and differentiation
- Public mentions in tech press
- Any red flags in company history
MARKET RESEARCH:
- Embedded finance market size and growth
- Key trends driving the market
- Regulatory environment
- Major players and recent M&A activity
COMPETITIVE POSITION:
- Direct competitors and their funding
- How does this company's approach differ?
- Defensibility of their position
RISK ASSESSMENT:
- Technology risks
- Market risks
- Team/execution risks
- Regulatory risks
DUE DILIGENCE QUESTIONS:
- What questions should I ask the founders?
- What proof points should I request?
- What would be deal-breakers?
The Verification System
Never trust AI research blindly. Use this verification system.
The 3-Layer Verification
LAYER 1: Source Check
├── Is the source cited actually authoritative?
├── Can I find the original source?
└── Is the information current?
LAYER 2: Cross-Reference
├── Do multiple independent sources agree?
├── Are there contradicting sources?
└── What explains any contradictions?
LAYER 3: Logic Check
├── Does the information make logical sense?
├── Are there obvious gaps or inconsistencies?
└── What questions remain unanswered?
Verification Prompt Template
For the research you just provided, I want to verify accuracy:
1. For the top 5 most important claims/statistics, provide:
- The exact source (document name, page if possible)
- Date of the source
- Alternative sources that confirm this
- Your confidence level (1-10)
2. Are there any claims where:
- You're uncertain about the source?
- Sources conflicted and you chose one?
- The information might be outdated?
3. What would change your conclusions if:
- Recent data (last 30 days) contradicts this?
- I'm in a different geographic market?
- My specific context differs from the general case?
Red Flags to Watch For
Templates & Cheatsheets
Quick Research Prompt Templates
Market Research:
Research [market/industry] comprehensively:
- Market size and growth (with sources)
- Key players and market share
- Trends driving growth
- Challenges and barriers
- Future outlook (2-3 years)
- Entry opportunities
Target audience: [role] at [company type]
Required sources: Industry analysts, recent reports (2024+)
Technology Evaluation:
Evaluate [technology/tool] for [specific use case]:
- Core capabilities
- Limitations and constraints
- Comparison with alternatives
- Real-world implementation examples
- Total cost of ownership
- Implementation complexity
- Recommendation with reasoning
Technical depth: [Beginner/Intermediate/Advanced]
Trend Analysis:
Analyze the trend of [trend] in [industry]:
- Current state and adoption
- Driving factors
- Barriers to adoption
- Key players/innovators
- Predictions for next 2 years
- Action items for [specific role]
Focus on practical implications, not hype.
Research Quality Checklist
## Research Quality Checklist
### Completeness
- [ ] All questions answered
- [ ] Sufficient depth for decision-making
- [ ] No obvious gaps
### Accuracy
- [ ] Sources cited for key claims
- [ ] Sources are authoritative
- [ ] Information is current (within 12 months)
### Balance
- [ ] Multiple perspectives included
- [ ] Counter-arguments addressed
- [ ] Limitations acknowledged
### Actionability
- [ ] Clear conclusions drawn
- [ ] Recommendations provided
- [ ] Next steps identified
### Confidence
- [ ] Uncertainty acknowledged where relevant
- [ ] Confidence levels indicated
- [ ] Areas for further research noted
The 10 Commandments of Deep Research
- Be Specific – Vague questions get vague answers
- State Your Context – Your situation affects what's relevant
- Request Structure – Define how you want information organized
- Demand Sources – Always ask for citations
- Iterate – First results are starting points, not endpoints
- Verify – Cross-check important claims
- Challenge – Ask for counter-arguments
- Quantify – Request numbers, not just descriptions
- Time-Bound – Specify recency requirements
- Act – Research without action is just reading
Conclusion
Gemini Deep Research is the most powerful research tool most people have ever had access to. But like any powerful tool, it requires skill to wield effectively.
Master the PRECISE framework. Use the iteration loops. Verify your findings. And most importantly—be specific about what you need.
The researchers who master this tool will have an unfair advantage over those who don't.
Now go research something that matters.
The best research doesn't just find information—it transforms it into actionable insight.
Tags: #Gemini #DeepResearch #AI #Research #Productivity #GoogleAI #ResearchTips #KnowledgeManagement #AITools #DataResearch
Comments · 0
Beta: comments are stored locally on your device and not visible to other readers.
No comments yet. Be the first to share your thoughts.