rsa-strength-copy-diversity-audit

name

rsa-strength-copy-diversity-audit

description

For each Google Ads ad group, judge whether its Responsive Search Ads are built to win the auction — strong Ad Strength, diverse non-redundant headlines, and pinning that doesn't strangle rotation. Use this skill whenever the user audits a Google Ads account, mentions Ad Strength, responsive search ads, RSAs, ad copy, headlines, pinned headlines, duplicate / near-identical headlines, "why is my Ad Strength Poor", or asks which ads to rewrite — even if they don't say "ad copy". Judges the AD COPY ITSELF (strength + headlines + pinning) ONLY; ad EXTENSIONS/assets belong to the complementary `ad-extensions-assets-audit` skill, and message-match to the keyword/landing belongs to the `keyword-ad-landing` cluster.

Ad Copy — RSA Strength & Headline Diversity

Goal (job-to-be-done)

Per ad group, answer the client's question: are the Responsive Search Ads strong enough to win the auction? Read each RSA's Ad Strength rating, judge whether its headlines are diverse (not near-duplicates that risk serving two near-identical lines together), and flag pinning that limits Google's rotation. The unit of analysis is the ad (inside its ad group); the output is a strength/diversity verdict + the exact headlines to fix or unpin.

• Who: media buyer / PPC manager. When: onboarding audit + periodic creative hygiene.
• Decision it drives: which ads to rewrite, which headlines to replace (LOW labels) or keep (BEST), which pins to remove so Google can rotate.
• The differentiator: it doesn't just report the Ad Strength chip — it explains why it's weak (over-pinning, near-duplicate headlines, too few assets) and names the exact headline to act on, using Google's own action_items as supporting evidence.

Scope

• ✅ Ad Strength rating (POOR/AVERAGE/GOOD/EXCELLENT), headline diversity / near-duplicates, pinning that limits rotation, per-asset performance labels when populated.
• ❌ Ad extensions / assets presence (sitelinks, callouts, snippets, images) → complementary ad-extensions-assets-audit.
• ❌ Message-match to the keyword / landing page (does the copy deliver the search's promise?) → the keyword-ad-landing cluster.
• ❌ Creative angle ideation (what new hook to write) — that's open-ended judgment, not this audit.

Components (read these references as needed)

• Tools / data plan: references/tools.md — the ad-grain query: headlines (text + label + pin), Ad Strength, action items, impressions.
• Framework / rubric: references/framework.md — the brain: the strength distribution, the pinning/duplicate/few-assets logic, the PENDING-label trap, when it does NOT apply.
• Output schema: references/output.md — the JSON this skill emits.

Operate

Input: per ad — ad_strength, the list of responsive_search_ad_headlines (each with text, assetPerformanceLabel, pinnedField), action_items (Google's own ad recommendations), and impressions to rank by serving volume. No business context required beyond the account itself.

Process: apply the rubric in references/framework.md. For each ad, sorted by impressions desc: read the Ad Strength; flag POOR/AVERAGE. Scan the headline list for near-duplicates (same root reworded — "X Software" / "X App") and for pinning (esp. a brand pinned to HEADLINE_1, which limits rotation and drags strength). Count headlines — too few assets is its own flag. Read assetPerformanceLabel: when PENDING, say "not enough data yet"; when populated, LOW = replace-candidate, BEST = keep. Use action_items as supporting evidence, never the sole verdict.

Emit the JSON in references/output.md:

• synthesis — the canonical three strings: headline (the weakest high-impression ad + the one move), diagnosis (the account's dominant weakness — over-pinning vs duplicate headlines vs thin assets), action (the highest-leverage ad to fix, where / what / why).
• ads[] — one per RSA: strength, headline_count, issues[], a verdict, and a recommendation {where, what, why}.

A renderer (the orchestrator's formats/*) turns this JSON into the human document. Emit pure data — no emojis, tables, markdown, or colors in the output.

Voice (don't copy the rules, link them): write every narrative line per ../../../_framework/writing.md — the question the data answers yes/no; the metric/label carried as data, never spelled out in prose; first sentence answers the heading, then names the driver. Plain language for a non-technical owner ("unpin the brand headline so Google can test more combinations"), the technical setting in parentheses.

Example (illustrative — NOT rules)

• Over-pinned: an RSA with Ad Strength AVERAGE and the brand pinned to HEADLINE_1 → over_pinned → unpin_headline: the pin blocks Google from rotating, capping strength.
• Duplicate headlines: two headlines "Acme Insurance Software" / "Acme Insurance App" that can serve together → duplicate_headlines → replace_headline: rewrite one to a distinct benefit.
• Thin + pending labels: an ad with only 4 headlines, all assetPerformanceLabel: PENDING → few_assets → add_headlines; don't call any single headline weak yet — there isn't enough data.