Prompts to build a customer knowledge center

Prompt 1: Surface customer language for copy

• Best for: Finding the exact words and phrases your customers use to describe your product or process
• Use with: ChatGPT Projects / Claude Projects

Variables to fill in:

• [SPECIFIC PROCESS OR FEATURE]—e.g., ​“our onboarding process,” ​“the payment experience,” ​“customer support response time”

What to expect: A paragraph summarizing language patterns, followed by a list of actual customer phrases with citations. Use these directly in landing page copy, email subject lines, or ad creative.

Follow-up prompt: 

“Now write 5 headline options for a landing page about [FEATURE] using only language from these customer quotes.”

Prompt 2: Find targeted testimonials

• Best for: Pulling testimonials that match a specific audience segment, location, or use case
• Use with: ChatGPT Projects / Claude Projects

Variables to fill in:

• [NUMBER]—e.g., ​“3–5”
• [SEGMENT DESCRIPTION]—e.g., ​“at mid-market companies,” ​“in the East Coast region,” ​“using our enterprise plan”
• [CHARACTER LIMIT]—e.g., ​“280” for social media, ​“150” for display ads

What to expect: Edited testimonials ready to drop into campaigns, each attributed and length-constrained. Review for accuracy—AI may trim context that changes meaning.

Prompt 3: Extract constructive criticism by segment

• Best for: Understanding pain points from a specific customer group for product or messaging improvements
• Use with: ChatGPT Projects / Claude Projects

Variables to fill in:

• [CUSTOMER SEGMENT]—e.g., ​“mid-market,” ​“new customers (first 90 days),” ​“enterprise”
• [SPECIFIC TOPIC]—e.g., ​“our latest software feature,” ​“the onboarding experience,” ​“pricing”

What to expect: Feedback organized by theme (e.g., ​“complexity,” ​“missing features,” ​“pricing confusion”) with direct quotes and messaging suggestions. Use this to preemptively address objections in campaigns.

Prompt 4: Create a customer persona from real data

• Best for: Building a customer persona grounded in actual feedback rather than assumptions
• Use with: ChatGPT Projects / Claude Projects

Variables to fill in:

• [PERSONA TYPE]—e.g., ​“our most satisfied customers,” ​“customers who churned,” ​“power users”
• [PRODUCT/SERVICE]—your product or service name

What to expect: A data-backed persona with actual customer language, not marketing-speak. Compare against your existing personas to see what’s different.

Follow-up prompt: 

“Now act as this persona. I’m going to describe a new campaign concept, and I want you to respond as this customer would, including what excites them, what concerns them, and whether they’d engage.”

Prompt 5: Deep research for strategic analysis

• Best for: Quarterly reviews, strategy sessions, or identifying overlooked opportunities
• Use with: ChatGPT Deep Research / Claude Extended Thinking

Variables to fill in:

• [TIME PERIOD]—e.g., ​“quarter,” ​“6 months,” ​“year”
• [MARKETING/PRODUCT]—e.g., ​“email marketing,” ​“product roadmap,” ​“content strategy”

What to expect: A multi-page research report suitable for sharing with stakeholders. Best used as a starting point for deeper strategic analysis, not as a final deliverable.

Tips for better results

• Always require citations: add ​“Cite all quotations and customer feedback with the source document and name of the customer” to your custom instructions. This prevents hallucinations from making it into your marketing.
• Upload regularly: your knowledge center is only as good as its data. Add new call transcripts, reviews, and survey responses monthly.
• Include campaign results: upload post-mortem notes and metrics so the AI can connect customer sentiment to campaign performance.
• Start specific, then go broad: query about one feature or segment first. Once you trust the output, try broader strategic questions.