skills/ad-platform/references/rejection-taxonomy.md

# The rejection taxonomy — why converting copy gets banned

A batch of high-performing ad texts (~20 creatives across 12 languages) was
rejected by Meta for **"fraud, scams and deceptive practices."** The ads
worked commercially — the emotional pattern was right. The problem was a
specific set of *language and format choices* layered on top of a sound
emotional core. This file separates the two: keep the emotional core, replace
the risky language.

Examples use the fictional placeholder **VELORA** — they illustrate structure,
never product scope.

## The recurring rejected structures

1. **Checklist / "how it works" funnel ending in a guaranteed outcome**:
   "Upload your product once → AI finds your ideal customers → creates ads
   that grab attention → runs campaigns everywhere → sales start coming in →
   your product goes viral."
2. **Hyper-specific, unverifiable result stats**: "Sell 1,000 this week",
   "190+ sold this week alone", "43 orders before lunch", day-by-day
   escalating numbers (Monday: 4, Wednesday: 11, Friday: 23…).
3. **Absolute "never again" claims**: "Never manage your ads manually again."
4. **An extreme low price as the entire hook**: "Use an AI agent for $1 a
   day" — a price the advertiser really charges is still flagged when it is
   stated as an absolute and made to carry the whole ad.
5. **Fabricated urgency**: "Less than 24 hours left", "70% discount ends
   tonight" — running indefinitely across audiences and dates, tied to no
   real promotion window.
6. **Street-flyer / handwritten-note testimonial format**: a torn paper note
   pinned to a pole, a first-person income story, granular daily numbers.
   The visual trope is itself scam-coded, independent of the copy.
7. **POV meme with an implausible speed/effort ratio**: "your business
   partner is still struggling in Canva… meanwhile you added ONE photo and
   had 43 orders before lunch."

## Why each pattern gets flagged

The policy targets ads promising atypical, guaranteed, or unverifiable
business/income outcomes, and formats that mimic classic scam creative:

| Pattern | Why it reads as scammy to a reviewer |
|---|---|
| Guaranteed outcome ending a checklist | Implies a guaranteed business result from minimal effort — the get-rich-quick shape, regardless of how true it is for some users. |
| Exact day-by-day / hour-by-hour stats | Reads as fabricated precision. Nobody can verify it — implausibly specific numbers are a *stronger* red flag than vague ones. |
| "Never again" / absolute claims | Absolute performance claims are treated as unsubstantiated guarantees. |
| Extreme low price as the entire hook | Pairing an extreme price with an extreme outcome reads as bait, even when the price is accurate. |
| Evergreen fake urgency | Scarcity not tied to a real, single, time-bound promotion is explicitly deceptive-practice territory. |
| Handwritten flyer visual + income story | The visual format is a known scam-ad trope; it stacks with the copy risk. |
| POV meme, implausible speed/effort ratio | Same as the first row — a guarantee with near-zero effort. |

None of these are about the *product*. They are about **absolutes,
unverifiable specificity, guaranteed outcomes, and scam-coded formats.**
That is the line to walk: keep the appeal, lose those four things.

A neighbouring trigger class worth knowing: copy framed around the reader's
**advertising spend or agency fees** ("pause your ad budget", "fire your
$2,000/month agency") can be reviewed under financial-services policies as
well as deceptive-practices ones. The stack of money framing + outcome claim
is the highest-risk shape in this taxonomy.

## Rewriting inside the rules — worked shapes

Each keeps the emotional shape that converted, with the risk removed:

**Absolute → reduction** (was "Never do X again"):
- "Tired of doing X yourself? Let VELORA take the busywork off your plate."
- "Spend less time on X, more time running your business."
- "Built for people who'd rather not open a content calendar every morning."

**Guaranteed cascade → capability steps** (was "…→ sales pour in → viral"):
- End the checklist on a feature or a relief: "One less thing to remember
  every day." / "No more guessing what to post next." / "You get your time
  back." Never on an outcome.

**Implausible POV → relatable POV**:
- "POV: your co-founder is still fighting with Canva, and your week's posts
  are already scheduled." (The contrast is effort, not a revenue claim.)

**Fabricated income story → modest testimonial**:
- "I used to spend hours on posts that went nowhere. VELORA changed that for
  me." — outcome modest, plausible, never itemized day by day.

**Evergreen urgency → real, dated offer**:
- "This week only: try VELORA at a lower price." — used only while a real
  promotion runs; otherwise lead with the benefit and drop time pressure.

## Re-testing a rejected line

A rejection is sometimes appealed and overturned; a line that cleared appeal
once is **not** pre-cleared for the next submission, account, or reviewer.
When the user knowingly re-tests a risky line, build what was asked, name the
specific pattern it matches, and budget for a rejection-and-appeal cycle in
the plan rather than assuming first-pass approval. The record of any such
history is brand evidence — the brand's `history/` files, not this skill.