Skip to content
Academy
Marketing Academy · Field Work●AI in Marketing
MiniTeardown· 25 minutes

Prompt Teardown: Diagnosing 3 Broken Marketing Prompts

Duolingo

Objective: Given 3 flawed marketing prompts that reliably produce hallucinated, generic, or unusable marketing copy, apply the RCTF framework, constraint rules, and few-shot principles to diagnose every missing component and rewrite them into high-performing production briefs.

You are the Senior Growth Marketing Manager at Duolingo reviewing prompt templates submitted by regional marketing teams for push notifications, reactivation emails, and paid ad creative. Many prompts are yielding generic, cliché-ridden copy that sounds like a corporate bank rather than Duolingo's iconic, cheeky, slightly unhinged owl persona. You need to teardown and fix these prompts.

Analyze 3 defective prompts. Identify all missing RCTF components, absent brand constraints, ungrounded zero-shot risks, and provide actionable fixes.

Before you start

What you'll need

Free path (everything below is enough to finish)

FreePrompt defect scoring and rewrite worksheet

Simple collaborative template for redlining prompt briefs

FreemiumTesting flawed vs repaired prompt outputs

Direct side-by-side output comparison in free chat interface

The process

Specimens to review

Teardown this push notification prompt. Identify why it fails to produce on-brand Duolingo copy and name every missing structural component.

Sample output
Write 5 push notifications to get people to practice French on Duolingo today. Make them catchy, fun, and urgent so users click. Include emojis.

Specimen: synthetic, realistic

Teardown this comprehensive campaign prompt. Identify all structural flaws and failure modes.

Sample output
You are a world-class marketing genius. Write our entire Q3 back-to-school marketing campaign for Duolingo for Schools. We need a landing page headline and subheadline, a 5-part email nurture sequence for high school Spanish teachers, 10 Google Ads headlines with descriptions, 3 TikTok script concepts with viral hooks, and a press release announcing our new gamified teacher dashboard. Make it viral, professional, and conversion-optimized.

Specimen: synthetic, realistic

Teardown this market research prompt. Identify why this prompt produces hallucinated citations and subjective fluff instead of structured competitive intelligence.

Sample output
Compare Duolingo vs Babbel vs Rosetta Stone for adult language learners. Which app is the best and why? Write 3 paragraphs explaining their pricing and features.

Specimen: synthetic, realistic

Final deliverable

A 3-prompt teardown score sheet detailing missing RCTF components, constraint violations, and complete rewritten production prompt templates.

See a reference example
Sample output
Babbel — Prompt Teardown & Repair Matrix (Excerpt)

DEFECTIVE PROMPT:
  'Write a Facebook ad for Babbel German course targeting travelers. Make it good.'
DIAGNOSIS:
  Missing Role: No expert copywriter persona defined.
  Missing Context: Fails to specify traveler use-case (ordering food, airport navigation, conversational confidence in 3 weeks).
  Missing Constraints: No 125-char primary text limit; zero negative word exclusions.
REPAIRED RCTF PROMPT:
  [ROLE] You are a direct-response paid social copywriter specializing in adult language learning apps.
  [CONTEXT] Audience: English speakers traveling to Germany/Austria in 30 days. Value prop: 15-minute conversational lessons focused on real-world travel dialogues.
  [TASK] Write 3 primary text options (under 125 characters) and 3 headline options (under 27 characters).
  [CONSTRAINTS] No exclamation marks. Do not use 'fluent' or 'master'. Focus on ordering food and asking for directions.
  [FORMAT] Output in a 2-column Markdown table with columns: Asset Type, Copy Text.

Success criteria

You're done when you can:

  • Correctly identifies all missing RCTF components across the 3 flawed specimens
  • Diagnoses over-prompting and explains the attention context breakdown mechanism
  • Provides actionable, grounded prompt repairs following negative constraint and few-shot rules