thinkhard.SI

Prompt repository Marketing

Find the words customers use in reviews and feedback

You want headlines and copy built from how customers actually describe their problem and your product, taken from reviews, surveys or support messages.

Registry no.
MKT-027
Collection
Marketing
Laid out for
Claude
Score
10 out of 10
Blanks to fill
5
Added

MarketingClaude

The prompt

Score 10 of 10, Excellent

Use it when You want headlines and copy built from how customers actually describe their problem and your product, taken from reviews, surveys or support messages.

  • {{feedback}}
  • {{product}}
  • {{source}}
  • {{team}}
  • {{copy_use}}
Full prompt and why it works
Prompt
<customer_feedback>
{{feedback}}
</customer_feedback>

<context>
What we sell: {{product}}
Where this feedback comes from, and how much there is: {{source}}
</context>

Read the customer feedback above and pull out the exact words customers use, so our {{team}} can write copy for {{copy_use}} in the customers' language rather than ours. Copy works when readers recognize their own words in it, so quotes matter more than summaries here. A good result is one a copywriter can use without rereading the feedback, with every phrase traceable to a customer.

Format: the five numbered sections below, each under its name as a heading.
1. Themes: 4 to 8, each with a short label and the number of separate pieces of feedback that mention it, most frequent first. Cover the problem before buying, what made them choose, what they get from it, and their doubts or complaints.
2. Quotes: for each theme, 3 to 5 verbatim quotes (all there are for a thin-evidence theme), copied word for word, each with its source or line number so it can be traced.
3. Phrases worth reusing: 10 to 15 short phrases in customers' own wording, each with its theme and how many customers use it.
4. Doubts to answer: hesitations and objections customers mention, in their words, with a count for each.
5. Copy starters: 5 headline ideas built from the phrases above, each followed in brackets by the quote it comes from.

<example>
The form of one quote only: "I stopped dreading Monday payroll" (review 14, theme: time saved)
</example>

Every quote must appear in the feedback exactly as written. Do not tidy, combine or invent quotes, and do not count one customer twice. If a theme rests on only one or two pieces of feedback, label it "thin evidence". If the feedback is too short or too mixed to support themes, say so and list what you can find without forcing a pattern. Leave out customers' names and personal details. This is for internal use; before any quote appears in published marketing, I will check we have permission to use it.

Variables

{{feedback}}
Reviews, survey answers or support messages, one per line and numbered if you can, with personal details removed.
{{product}}
What you sell, in a sentence.
{{source}}
Where the feedback comes from and roughly how many pieces, for example "212 app store reviews from 2025 and 2026".
{{team}}
Who will use the findings, for example "marketing team".
{{copy_use}}
What the copy is for, for example "the new homepage and two ad campaigns".

Why it works

  • Background facts. The feedback sits in its own tag above the instructions, and every quote carries its line number, so each phrase can be traced back to a real customer.
  • What to avoid. Quotes must be word for word and never tidied or combined, because a phrase the model polished is no longer the customer's language.
  • If unsure. Themes resting on one or two comments are labeled thin evidence, so a single loud review does not become your headline.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/customer-language-from-reviews

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "customer-language-from-reviews"}, or just ask for it by name.