Pass or Fail: Teardown of Three AI-Generated Marketing Outputs
Objective: Given three synthetic AI-generated marketing outputs (a chatbot response, a testimonial ad, and a product image caption) from a small finance NBFC, apply the lesson's Human Review Gate checklist to identify which specific defects each output has, distinguishing real violations from plausible-but-fine content.
You're the content-review lead at Five-Star Business Finance, the Chennai-founded, NSE-and-BSE-listed MSME lender. Three AI-generated outputs are queued for publish tomorrow and you're the last check before they go live.
Run each specimen through the lesson's Stage 3 review checklist (sourced statistics, verified claims, legal sign-off triggers, copyright signals, tone-and-context fit), separate real defects from things that only look risky, and flag severity.
Before you start
What you'll need
Free path (everything below is enough to finish)
A simple tracking sheet is enough to document a 3-item review pass
The process
Specimens to review
This chatbot response is queued to go live on the website widget. Run it through the Stage 3 checklist. What defect(s) does it have, and how severe are they?
Customer support chatbot response: "Great news! Based on your profile, you're pre-approved for our MSME growth loan at 9.5% interest with zero processing fee. This offer is guaranteed and won't change regardless of your final documentation."
Specimen: synthetic, realistic
This testimonial ad is scheduled for a paid social campaign. What defect(s) does it have under Stage 4's disclosure rules?
Social ad copy: "'Five-Star Finance got me my loan in 24 hours, no paperwork hassle at all!' - Rajesh K., Coimbatore" (No disclosure that the testimonial was AI-generated based on aggregated customer feedback themes, not an actual quoted customer.)
Specimen: synthetic, realistic
This blog excerpt is also queued to publish. Does it pass the Stage 3/4 checklist, or does it have a defect?
Blog post excerpt (AI-drafted, human-edited): "MSME loan applications typically require GST registration, 2 years of business vintage, and recent bank statements. Processing times vary by lender and document completeness; check with your relationship manager for your specific timeline." A byline notes: "Drafted with AI assistance, reviewed and fact-checked by the Five-Star editorial team, August 2026."
Specimen: synthetic, realistic
Final deliverable
A pass/fail verdict for each of the 3 specimens with the specific defect, severity, and lesson-linked reason cited.
See a reference example
Go Digit General Insurance, AI-output review log (excerpt) Specimen: Claims chatbot response guaranteeing a payout amount before adjuster review Verdict: FAIL, critical Defect: States a specific payout figure as guaranteed before assessment is complete, mirrors the Air Canada liability pattern Specimen: AI-drafted policy-comparison blog post, disclosed AI-assistance byline, hedged claims Verdict: PASS Note: Disclosure present, claims appropriately hedged, no fabricated statistics
Success criteria
You're done when you can:
- Correctly identifies the chatbot's guaranteed-rate claim as a critical liability defect
- Correctly identifies the undisclosed AI testimonial as a disclosure violation
- Correctly passes the third specimen instead of over-flagging appropriately hedged, disclosed content