Persona Research Prompt Templates for Claude

Persona research is one of the tasks where Claude genuinely earns its keep — but only if the prompts are structured to extract specific, usable information rather than generic audience summaries. The templates below were developed and tested against Claude 3.5 Sonnet and refined through repeated use on B2B SaaS, e-commerce, and agency research workflows. Each one targets a distinct phase of persona research: discovery, psychographic depth, pain point extraction, and validation framing.

What makes persona prompts work differently on Claude

Claude handles persona research differently than GPT-4 in a few practical ways. It tends to be more cautious about making demographic claims without stated evidence, which is actually useful — it surfaces assumptions you didn't know you were making. It also responds well to role-framing ("You are a qualitative researcher...") and handles long context windows reliably, so you can paste in raw customer interview transcripts and ask it to extract persona signals from the text directly.

The main failure mode is over-hedging. Without specific constraints in the prompt, Claude will often produce persona outputs that read like "some users may prefer X while others prefer Y" — technically accurate but operationally useless. The templates below are written to push past that by forcing specificity through structured output formats and explicit scope constraints.

Template 1: Initial audience segmentation

Use this when you're starting from a product description or value proposition and need Claude to identify distinct audience segments worth researching further. It works best for B2B products with multiple buyer types or e-commerce categories with divergent shopper motivations.

You are a qualitative market researcher specializing in audience segmentation.

Product/service: [PASTE PRODUCT DESCRIPTION OR VALUE PROPOSITION]

Task: Identify 3–5 distinct audience segments who would plausibly purchase or use this product for meaningfully different reasons. For each segment:

1. Give the segment a descriptive name (not a demographic label like "35–44 females" — use a behavioral or motivational label like "Deadline-Driven Freelancer")
2. State the primary job-to-be-done this segment is hiring the product for
3. List the 2–3 alternative solutions they currently use or have tried
4. Identify the single biggest friction point with those alternatives
5. Describe what a successful outcome looks like to this segment (be specific — avoid "saves time" without quantifying what that means to them)

Do not combine segments that have meaningfully different buying triggers. If you're uncertain whether two segments should be split, err toward splitting and explain why.

Known failure modes

• Claude may default to industry-standard persona archetypes ("the budget-conscious buyer", "the power user") if the product description is too generic. Feed it specific features, pricing tier, or customer quotes to get more differentiated output.
• If your product description includes marketing language ("all-in-one solution", "seamless workflow"), Claude tends to mirror that vagueness back. Strip superlatives before pasting.
• Segment names sometimes drift toward demographic labels despite the instruction. If this happens, add: "Segment names must describe behavior or motivation, not age, gender, or job title."

Template 2: Psychographic depth interview simulation

This template is useful when you have a segment defined but need to develop the psychological and motivational texture — the kind of detail that makes messaging feel like it was written for a real person. It simulates a qualitative interview rather than asking Claude to summarize a persona.

You are playing the role of [SEGMENT NAME] — [1–2 sentence description of who they are and what they do].

I'm going to ask you a series of questions. Answer as this person would in a candid conversation, not as a marketing persona. Use first-person. Be specific about tools, time, money, and frustrations. Do not give aspirational answers — give honest ones.

Questions:
1. Walk me through the last time you had to solve [CORE PROBLEM THIS PRODUCT ADDRESSES]. What did you actually do, step by step?
2. What did you try first? Why didn't it work well enough?
3. When you're evaluating whether a new solution is worth trying, what's the first thing you check?
4. What would make you immediately close a tab and not come back?
5. If this product worked exactly as promised, what would change about your week?
6. Who else in your organization (or life) would be affected by this decision, and would they push back?

After answering all questions, step out of the persona and flag: (a) any answers where you were speculating beyond what the segment description supports, and (b) any questions where the answer would likely vary significantly within this segment.

The final instruction — asking Claude to flag speculation — is what separates this from a creative writing exercise. It forces a meta-layer that keeps outputs usable for actual strategy work rather than just inspiration.

What to adjust

• Replace the six questions with your actual research gaps. If you already know the evaluation criteria, swap question 3 for something about messaging — "What's the last piece of marketing copy that made you actually click?"
• For B2C personas, remove question 6 (organizational pushback) and replace with something about social context: "Would you tell anyone about this purchase? Why or why not?"
• The more detailed your segment description, the less Claude speculates. Include job title, company size, current tools, and a sentence about their relationship to the problem.

Template 3: Pain point extraction from raw text

When you have real source material — customer support tickets, review scrapes, interview transcripts, forum threads — this template structures Claude's extraction into a usable format. Claude handles long text well, so you can paste in 3,000–5,000 words of raw input without significant degradation.

Below is raw text from [SOURCE TYPE: customer interviews / support tickets / app store reviews / community forum]. Read it carefully before responding.

[PASTE RAW TEXT]

---

Extract the following from this text:

1. PAIN POINTS: List every distinct frustration, complaint, or unmet need mentioned. Quote the source text directly for each one (even a partial quote). Do not paraphrase — use the customer's language.

2. TRIGGER MOMENTS: Identify any moments described where the person decided to look for a solution. What was happening right before that decision?

3. LANGUAGE PATTERNS: List 8–12 specific words or phrases customers use to describe the problem. These should be verbatim from the text, not your summaries.

4. IMPLICIT EXPECTATIONS: What did customers expect that they didn't get? These are often stated as disappointments ("I thought it would...", "I assumed...").

5. GAPS IN THE TEXT: What questions about this persona can't be answered from this source material alone?

Format each section as a numbered list. Do not add interpretive commentary within the lists — save that for a separate section labeled ANALYST NOTES at the end.

Template 4: Competitive displacement framing

This one is specifically for understanding why a persona would switch from a competitor — or why they wouldn't. It's useful for positioning work and for identifying the messaging angles that actually move people versus the ones that sound good internally.

Context:
- Our product: [BRIEF DESCRIPTION]
- Competitor they currently use: [COMPETITOR NAME AND CATEGORY]
- Target persona: [SEGMENT NAME AND DESCRIPTION]

Task: Analyze the switching decision this persona faces. Structure your response as follows:

STAYING FORCES (reasons they won't switch even if our product is objectively better):
- List 4–6 specific, concrete reasons. Include switching costs, habit, integration lock-in, social/team dynamics, and risk aversion.

SWITCHING TRIGGERS (what would actually cause them to evaluate alternatives):
- List 3–5 specific events or moments that would prompt this persona to look. These should be situational triggers, not feature comparisons.

EVALUATION CRITERIA (what they'd use to compare options once they're looking):
- List the 5–7 criteria in priority order, as this persona would rank them — not as we would rank them.

MESSAGING THAT BACKFIRES:
- List 3–4 common marketing claims that would actually increase skepticism or distrust for this persona. Explain why each one backfires.

Be specific. If you're making assumptions about the competitor's product, flag them.

Template 5: Persona validation checklist

Once you've built a persona document — whether from AI synthesis or actual research — this template asks Claude to stress-test it. The goal is to surface internal inconsistencies and gaps before you build messaging or content strategy on top of a flawed foundation.

Below is a persona document. Review it critically as a qualitative research methodologist.

[PASTE PERSONA DOCUMENT]

---

Evaluate this persona on the following dimensions and give a rating of STRONG / ADEQUATE / WEAK for each, with a 1–2 sentence explanation:

1. SPECIFICITY: Are the attributes specific enough to make a real decision with? ("Values efficiency" is weak; "Needs to cut task time by 30% to meet a quarterly headcount freeze" is strong.)

2. INTERNAL CONSISTENCY: Do the stated behaviors, goals, and frustrations actually fit together? Flag any contradictions.

3. EVIDENCE BASIS: Which claims are supported by stated research, and which appear to be assumptions? List the assumptions separately.

4. ACTIONABILITY: Could a copywriter use this to write a headline? Could a product manager use it to prioritize a feature? If not, what's missing?

5. DISTINCTIVENESS: Does this persona describe a real segment, or could it describe almost anyone? What makes this person different from adjacent segments?

End with a list of 3–5 specific research questions that would most improve this persona's reliability.

Template comparison: when to use which

Combining templates into a research workflow

These templates work best in sequence rather than in isolation. A practical workflow for a B2B product launch looks like this:

1. Run Template 1 to identify 3–4 candidate segments from your product brief.
2. Run Template 3 on any existing customer text (even a handful of support emails or G2 reviews) to ground the segments in real language.
3. Run Template 2 for the 1–2 segments that seem most strategically important, using the language patterns from Template 3 to enrich the segment description.
4. Run Template 4 if you're entering a market with an established competitor — this is where positioning decisions get made.
5. Compile the outputs into a persona document, then run Template 5 before handing it to anyone who'll act on it.

The full sequence takes 60–90 minutes in Claude for a single persona. That's faster than a round of stakeholder interviews, but it's not a replacement — it's a way to arrive at those interviews with sharper hypotheses and better questions.

Limitations to keep in mind

• Claude cannot access real-time market data, recent survey findings, or your CRM. The outputs reflect general patterns, not your specific customer base.
• Template 2 (simulation) is the highest-risk format for hallucination. Claude will generate plausible-sounding specifics ("She uses Notion and Asana together") that may not reflect your actual users. The speculation-flagging instruction helps, but doesn't eliminate this.
• Prompts that work well in Claude 3.5 Sonnet may produce noticeably different outputs in Claude 3 Haiku (faster and cheaper, but less nuanced on inferential tasks) or future model versions.
• If your product is in a highly regulated category (healthcare, finance, legal), Claude may be more conservative about making behavioral claims. You may need to explicitly note that you're doing marketing research, not clinical or legal analysis.