/meta-ad-policy-checker

name: meta-ad-policy-checker
description: Pre-flight policy check for Meta ads. Takes ad copy plus advertiser context, resolves and fetches the relevant Meta transparency-center policy pages at runtime, and returns a Pass / Fix Required / Block verdict with cited findings and rewrites.

Meta Ad Policy Checker

Meta disapproves ads for policy reasons constantly, and most disapprovals are preventable. The pattern is almost always the same: an advertiser (or an AI agent generating variants) writes copy that uses a phrase or implies a claim that violates Meta's published advertising standards. The fix is cheap if you catch it before submission, expensive if you don't — repeated disapprovals on the same account can throttle delivery or trigger account-level restrictions.

This skill is a pre-flight check. It reads ad copy, figures out which Meta policies apply, fetches the live policy text from Meta's transparency center, and returns a clear verdict with specific findings, citations, and rewrites. It does not hardcode policy rules — Meta updates those, and a static rule list goes stale. Instead, the skill uses Meta's own canonical policy pages as ground truth on every run.

**Core principle:** The skill provides the methodology — what to check, how to reason, what severity to assign. Meta provides the source of truth — the actual policy text. This separation is what keeps the skill correct as Meta's standards evolve.

When to Use

• Before launching any new Meta ad
• After generating ad copy variants (call this on each variant before showing or submitting)
• When auditing an existing live ad for policy risk
• When troubleshooting a recently disapproved ad
• Before pushing copy through any Meta MCP / API write tool

Phase 0: Intake

1. **Advertiser context** *(1–2 sentences)*

• What the business does
• What product / service / offer is being advertised
• Primary conversion goal

2. **Ad asset(s)**

• Headline(s)
• Primary text / body
• Description (if applicable)
• CTA button text
• Destination URL *(optional — enables a landing-page cross-check)*

3. **Special Ad Category** — declared by user: `None` / `Employment` / `Credit` / `Housing` / `Social Issues, Elections or Politics` *(this changes the applicable rules significantly)* 4. **Targeting summary** *(optional)* — useful for discrimination checks (age, gender, location exclusions) 5. **Visual description** *(optional)* — text-in-image is policy-relevant 6. **Mode** — `pre-flight` (block until clean) or `audit` (flag-only, used for already-live ads)

Phase 1: Determine Which Policies Apply

Reason about the ad before fetching. Most ads need 3–6 policy pages, not all 25. Always include the **baseline set**, then add **content-driven** pages based on what the ad mentions, then add **category-driven** pages based on declared Special Ad Category.

Treat the entries below as policy lookup keys, not URL slugs. Meta's Transparency Center URLs are nested under category paths and change over time, so never construct a flat policy URL directly from these labels.

Baseline (every ad)

• Community Standards
• Personal Attributes
• Sensational Content
• Misinformation
• Engagement Bait / Spam
• Grammar & Profanity
• Non-Functional Landing Pages

Content-driven (add when present)

| Triggers | Policy lookup keys | |---|---| | Income, earnings, payouts, "make money", specific dollar amounts | Personal financial requirements; unrealistic outcomes | | Health, weight loss, supplements, wellness claims | Health and wellness; before-and-after photos | | Targeting by age, gender, race, religion, nationality | Discriminatory practices | | Crypto, weapons, adult, drugs, alcohol, gambling, tobacco | Restricted content | | Political, social-issue, election content | Social issues, elections or politics | | Profanity, slurs, sensitive language | Profanity; inflammatory content | | Anything that implies tracking / scraping / circumvention | Circumventing systems |

Category-driven (add based on declared Special Ad Category)

| Special Ad Category | Add | |---|---| | Employment | Employment | | Credit | Credit | | Housing | Housing | | Social Issues, Elections or Politics | Social issues, elections or politics |

Output of Phase 1: an ordered list of policy lookup keys to resolve and fetch in Phase 2.

Phase 2: Fetch + Cache Live Policy Text

Fetch Meta's ad-standards index first:

https://transparency.meta.com/policies/ad-standards/

For each lookup key from Phase 1:

1. Resolve it from the index to Meta's canonical policy page URL. Prefer exact title matches, then closest title / category matches. 2. Fetch the resolved canonical URL. Do not build a URL by appending the lookup key directly to the ad-standards base path; those flat URLs 404 for many current policies because Meta nests policy pages under category paths. 3. Cache the lookup key → canonical URL → page text mapping.

**Caching rule:** in-memory for the current session only. A batch check of 10 ad variants should produce 3–6 policy-page fetches total, not 30–60. Skip persistent caching to avoid stale-cache bugs across sessions.

**Fallback:** if the index cannot be parsed or a resolved page 404s, use a site-restricted web search for the policy title on Meta's transparency domain and navigate to the closest-matching current policy. Log this as a "policy URL drift" note in the output so the lookup list can be updated.

Phase 3: Reason — Ad vs. Policy

For each fetched policy, walk through every element of the ad (headline, body, description, CTA, link, visual description if provided) and ask:

1. **Does any phrase, implication, or pattern in the ad match a restricted pattern in this policy's text?** 2. **What specific clause from the policy applies?** Pull the direct quote. 3. **What is the severity?**

Severity model — three levels:

| Severity | Definition | |---|---| | **Block** | Clear violation. Ad will almost certainly be disapproved. Do