How do I find the right AI personas in the public library?
The short answer: describe who you need in plain language ("25-35 year-old female office workers in Shanghai who drink coffee often") and the library returns a ranked list. Pick 5–10 that span different concerns — not 10 versions of the same person.
The public library in 30 seconds
The library has ~5,000 pre-built personas covering common demographics. Each persona has a fixed background (age, city, job, income, habits, attitudes, pain points) and answers consistently from that background in any interview or discussion.
You don't need to find the "perfect" persona. You need a group of personas that, taken together, represents the variety of users in your real audience.
Three ways to search
Keyword search. The default. Type a description in natural language:
"25-35 year-old female office workers in tier-1 cities who drink coffee 3+ times a week"
"Fitness enthusiasts focused on muscle building"
"Price-sensitive online shoppers"
What you describe can be demographics, behaviors, scenarios, or attitudes. Mix freely.
Tag filtering. Useful when you want to narrow by specific dimensions. Common tag categories:
| Dimension | Examples |
|---|---|
| Age | 18-25, 25-35, 35-45, 45-55, 55+ |
| Gender | Male, Female |
| Occupation | Product manager, developer, designer, student, stay-at-home parent |
| Income | 5-10K, 10-20K, 20-30K, 30K+ |
| Lifestyle | Health-conscious, efficiency-focused, social, homebody |
| Consumption | Price-sensitive, quality-first, brand-loyal, early adopter |
| Interests | Tech, beauty, fitness, food, travel, gaming |
Combine tags to narrow further.
Similarity search. Best when you have a real user profile and want AI-matched personas. Type a description (up to 200 words) of your target user. The library returns personas ranked by similarity, with a match score.
How many personas to pick
| Study type | Recommended count |
|---|---|
| 1-on-1 Interview | 5–10 |
| Focus Group discussion | 3–8 |
| Large-scale validation | 50–100 |
For a typical product or market study, 5–10 personas is the sweet spot. Less than 5 and you don't have enough variety. More than 10 and each persona's voice starts to get diluted.
Pick for variety, not similarity
The point of using multiple personas isn't to triangulate "what everyone thinks." It's to surface the different ways your audience might react. A good selection spans different concerns:
Example for testing sparkling coffee:
- 2 price-sensitive personas (different age, city, income)
- 2 quality-first personas (different age, city, income)
- 1 social-persona (different age, city, income)
- 1 health-conscious persona
- 1 skeptic
Avoid picking 5 personas who all care about the same thing. The interesting findings usually come from the disagreement between personas.
What if the library doesn't have what you need
Three fallbacks:
Widen the search. Drop some constraints and see if usable personas emerge from a broader match.
Use Scout. Run a Scout task on the demographic you can't find. AI will observe public social media discussions and generate 5–10 personas from what it sees. Takes 1–2 days but the personas are tailored to your actual gap.
Build from your own data. Upload your existing customer interviews, surveys, or user research transcripts. AI extracts personas from those. Takes 10 minutes per transcript.
Don't compromise a study by picking the closest library match when the fit is poor. The fix is to build the right persona, not to use a wrong one.
When to bookmark a persona
If you find a persona you'll want to use again — your typical user, your champion customer, your persistent skeptic — click "Add to My Library." It stays accessible without re-searching.
Bookmarked personas are organized by you: tags, notes, custom labels. Useful for repeated studies on the same product.
Quality of library personas
Library personas are stable, consistent, and cover most general research needs. For specialized audiences — your specific VIP customer, a niche community, a B2B decision-maker — Scout-generated or custom-uploaded personas fit better.
Practical patterns
Quick validation (1 hour). Pick 3 personas with the most opposing views, run a Discussion. You get the disagreement and consensus in one transcript.
Comprehensive study (half day). Pick 8 personas across 3 segments, run parallel Interviews. Combine transcripts in one report.
Quarterly tracker. Bookmark a Panel of 8–10 personas. Re-interview them every quarter with a new question. Track how the same personas' views shift over time.
Related: Persona Library, Scout, Panel, Interview, Discussion