Skip to content
Academy
Marketing Academy · Field Work●AI in Marketing
All projects·From: AI Image Generation for Social Content
MiniTeardown· 30 minutes

Teardown: Spot the Defects in a DoorDash AI-Generated Carousel

DoorDash

Objective: Given 3 synthetic AI-generated Instagram carousel slide specimens for a DoorDash promo, apply the lesson's text-accuracy, item-count, and icon/color consistency rules to identify real defects and reject plausible-looking non-issues.

You're a freelance social media designer contracted by DoorDash's regional marketing team to QA a batch of AI-generated promo carousel slides before they go live.

Review each specimen against the lesson's rules on text accuracy, item count, icon consistency, and color role assignment. Flag every real defect, rate its severity, tie it to the rule that prevents it, and do not flag the distractors.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets
FreeLog each specimen's defects, severity, and the fix instruction to feed back into the next prompt

Free, no account friction, and easy to hand off to the DoorDash marketing contact as a review memo

The process

Specimens to review

Review this slide specimen against the lesson's text-accuracy guidance. List every defect you can find, rate each by severity, and explain why it matters before checking the answer key.

Sample output
Slide 1 of 3 — DoorDash 'Weekend Deals' carousel

Headline (large, top-center): 'FREE DELVIERY ALL WEEKEND'
Subhead: 'On orders over $15'
Background: warm red-orange gradient with DoorDash logo bottom-right
Bottom band: '3 CITIES ONLY' badge in a white pill

Specimen: synthetic, realistic

The headline promises 5 items. Count what's actually rendered and decide whether this is a real defect.

Sample output
Slide 2 of 3 — Headline: '5 Ways to Save This Week'

Four numbered cards rendered below it: 01 Free delivery, 02 20% off first order, 03 Double DashPass points, 04 $5 off $20+. No fifth card is present.

Specimen: synthetic, realistic

Compare the 3 icons and the accent color usage against the lesson's global-icon-instruction and color-role rules.

Sample output
Slide 3 of 3 — 'Why DashPass' 3-icon row. Icon 1 (delivery truck) and icon 2 (clock) are flat 2px line-style icons in navy. Icon 3 (percent sign) is a glossy, filled icon with a drop shadow, rendered in the same lime-green accent color used elsewhere on the slide for the 'delivery' label text.

Specimen: synthetic, realistic

Final deliverable

A defect log: one row per specimen, defect description, severity, and the specific targeted-edit instruction to feed back into the next prompt.

See a reference example
Sample output
Airbnb 'Host Spotlight' carousel — QA log (excerpt)

Slide 1: CRITICAL — subhead reads 'Earn extra icnome' (should be 'income'). Fix: targeted single-word edit, do not reroll the layout.
Slide 2: MODERATE — headline promises '4 host tips', only 3 cards rendered. Fix: restate the count twice and pre-label items 01-04 in the next prompt.
Slide 3: COSMETIC — soft drop shadow on the host avatar photo. No action needed, matches the approved style block.

Success criteria

You're done when you can:

  • Correctly flags all 4 real defects across the 3 specimens and rates each by severity
  • Does not flag any of the 5 distractors as defects
  • Ties the fix for each defect to the specific lesson rule that prevents it