When should I use AI personas instead of real people?

Question Type

Product Q&A

---

User's Real Concern

"I have budget and time, why not just do research with real people? AI personas, no matter how good, are still fake, right?"

Core anxiety: Worried about choosing the wrong tool and wasting time and money.

---

Core Answer

It's not either/or, each has optimal scenarios.

AI personas don't "replace" real people, but are more efficient and controllable in specific scenarios.

Quick Decision Framework

---

AI Personas' 5 Best Scenarios

Scenario 1: Exploratory Research (Don't know the answer)

Applicable:

• New product concept testing: "Will users buy?"
• Market opportunity exploration: "Does this need exist?"
• Competitor analysis: "Why do users choose competitors?"

Why AI is better:

• ✅ Speed: 3-4 hours vs 2-4 weeks for real recruitment
• ✅ Cost: Token cost ¥50-100 vs real people ¥3000-7000
• ✅ Low trial cost: Can restart immediately if direction is wrong

Real case: A coffee brand wanted to launch "sparkling coffee" and was uncertain about market acceptance.

• AI solution: Completed 5 in-depth interviews in 3 hours, found "users don't want to take risks in afternoon energy boost scenario"
• Decision: Adjusted positioning to "breakfast trial scenario", then validated with real people

---

Scenario 2: Large-scale User Research (5-10+ people)

Applicable:

• Comparing segmented user groups
• A/B testing multiple options
• Reactions in different scenarios

Why AI is better:

• ✅ Parallel efficiency: Conduct 5-10 interviews simultaneously, impossible with real people
• ✅ Consistency: Stable interview quality, not affected by interviewer's state
• ✅ Repeatable: Can have the same set of personas participate in multiple studies

Comparison:

---

Scenario 3: Rapid Prototype Testing (Iteration validation)

Applicable:

• Feature priority ranking
• Pricing strategy testing
• Copy/design option comparison

Why AI is better:

• ✅ Fast iteration: Test today, adjust today, test again tomorrow
• ✅ Low trial cost: No worry about wasting real respondents' time
• ✅ Trackable: Same set of personas, see how reactions change

Real case: A SaaS product tested 3 pricing plans, used AI personas for 5 rounds:

• Round 1: Monthly ¥99 → Feedback "too expensive"
• Round 2: Monthly ¥49 → Feedback "cheap but worried about quality"
• Round 3: Monthly ¥79 (20% off annual) → Found optimal balance
• Total time: 8 hours, cost ¥200
• If using real people: At least 2 months, cost ¥15,000+

---

Scenario 4: Competitor Comparison Research

Applicable:

• "Why do users choose competitors?"
• "Where are our advantages?"
• "What are competitors' fatal flaws?"

Why AI is better:

• ✅ Real behavior data: AI personas based on real social media observation
• ✅ Unbiased expression: Not worried about offending people, more honest negative feedback
• ✅ Multi-brand comparison: Can have personas evaluate 5-10 competitors simultaneously

---

Scenario 5: Sensitive Topic Pre-research (Before real people)

Applicable:

• Price sensitivity testing
• Negative feedback collection
• Potential pain point exploration

Why AI first:

• ✅ No social pressure: AI personas more directly express negative opinions
• ✅ Safe exploration: Avoid awkwardness or conflict in real interviews
• ✅ Risk prediction: Discover potential "landmine" questions in advance

Strategy:

1. Use AI personas for pre-research, find sensitive points
2. Adjust interview approach and question design
3. Then validate with real people, avoid interview failure

---

Real Interviews' 5 Must-choose Scenarios

Scenario 1: High-stakes Critical Decisions

Must use real people:

• Brand repositioning
• Major product redesign
• Core business direction adjustment

Reason:

• Decision cost too high, can't rely on AI alone
• Need real people's emotional resonance and intuitive judgment

---

Scenario 2: Deep Emotional Connection

Must use real people:

• Brand emotional value exploration
• Lifestyle deep understanding
• Cultural and value insights

Reason:

• AI cannot fully simulate real people's emotional depth
• Need non-verbal information from real people (tone, pauses, hesitation)

---

Scenario 3: Complex Psychological Analysis

Must use real people:

• Psychological trauma and emotional barriers
• Deep anxiety and fear
• Complex interpersonal motivation

Reason:

• Need professional psychologists to participate
• AI lacks ability to handle complex emotions

---

Scenario 4: Innovation Breakthrough (Not validation)

Must use real people:

• Creating entirely new categories
• Revolutionary product concepts
• Cross-industry innovation fusion

Reason:

• Only real people can propose ideas "AI personas can't imagine"
• Need real people's divergent thinking and creative sparks

---

Scenario 5: Stakeholder Decision-making

Must use real people:

• B2B multiple decision-makers
• Internal cross-department collaboration
• Complex stakeholder scenarios

Reason:

• AI cannot simulate complex organizational politics and interest dynamics
• Need real people's on-site judgment and adaptability

---

Combination Strategy (Best Practice)

Golden Combination: AI Exploration + Real Validation

Execution Flow:

Phase 1 - AI Exploration (1-2 days):

1. Use AI personas to quickly test 5-10 directions
2. Identify 2-3 most promising directions
3. Discover potential risks and controversial points

Phase 2 - Real Validation (1-2 weeks):

1. Design real interviews targeting key points discovered by AI
2. Validate if AI insights truly exist
3. Supplement emotional and intuitive judgments AI lacks

Advantages:

• ✅ Save real interview time and cost (reduce 80% ineffective exploration)
• ✅ Improve real interview quality (more precise questions)
• ✅ Lower decision risk (AI exploration + real confirmation)

Real case: A fitness app launched "AI personal trainer" feature:

• Phase 1 - AI exploration: Found users worry "AI doesn't understand my physical condition"
• Phase 2 - Real validation: Designed real interviews focusing on this concern
• Result: Confirmed concern is real, adjusted product direction to "AI assistance + human guidance"

---

Decision Tree: 3 Steps to Choose the Right Tool

Step 1: How important is this decision?

• High-stakes critical decision → Real validation
• Exploratory research → AI first

---

Step 2: Your time and budget?

• Sufficient time, adequate budget → Real or AI+Real combination
• Tight time, limited budget → AI first

---

Step 3: What depth needed?

• Deep emotional insights → Real
• Behavior and attitude research → AI

---

Common Misconceptions

Misconception 1: "AI personas are cheap substitutes"

❌ Wrong understanding: Use AI personas because they're cheap ✅ Correct understanding: AI personas are more efficient and controllable in specific scenarios

Example:

• Large-scale research (5-10 parallel): AI isn't cheap, it's impossible with real people
• Rapid iteration testing: AI isn't cheap, real people are too slow

---

Misconception 2: "AI personas will never match real people"

❌ Wrong understanding: Real people are better in all scenarios ✅ Correct understanding: AI personas achieve 85% consistency (exceeds human baseline 81%), surpassing human 81% baseline in some areas

Facts:

• Exploratory research: AI personas > real people (speed, cost, controllability)
• Deep emotional insights: Real people > AI personas (emotional resonance, non-verbal information)

---

Misconception 3: "Using AI means don't need real people anymore"

❌ Wrong understanding: AI can completely replace real people ✅ Correct understanding: AI and real people are complementary, not replacement

Best practice:

• Use AI for quick exploration, eliminate ineffective directions
• Use real people for deep validation, ensure correct decisions

---

Last Word

"AI personas aren't cheap substitutes, they're optimal solutions in specific scenarios. The key isn't 'AI vs real people', but 'when to use which'."

Remember:

• ✅ Exploratory research, large-scale surveys, rapid iteration → AI first
• ✅ High-stakes decisions, deep emotion, innovation breakthrough → Real must-choose
• ✅ Best practice: AI exploration + real validation

---

Related Features: Interview vs Discussion Document Version: v2.1