Google Ads RSAs

Read first:

• google-ads/references/operator-thesis.md
• google-ads/references/query-patterns.md
• google-ads/references/intent-map.md
• google-ads/references/rsa-playbook.md
• google-ads/references/deliverable-templates.md

Read workspace if available:

• workspace/ads/account.md
• workspace/ads/goals.md
• workspace/ads/intent-map.md
• workspace/ads/queries.md
• workspace/ads/winners.md
• workspace/ads/assets.md
• workspace/ads/learnings.md

Data Acquisition

Connected Mode (MCP available)

Pull via the search tool on google-ads-mcp:

Primary: RSA asset performance:

SELECT
  asset.text_asset.text,
  asset.type,
  ad_group_ad_asset_view.performance_label,
  ad_group_ad_asset_view.field_type,
  campaign.name,
  ad_group.name,
  metrics.impressions,
  metrics.clicks,
  metrics.conversions
FROM ad_group_ad_asset_view
WHERE segments.date DURING LAST_30_DAYS
  AND campaign.status = 'ENABLED'
ORDER BY metrics.impressions DESC

Supplementary: Search terms (for buyer language extraction):

SELECT
  search_term_view.search_term,
  campaign.name,
  metrics.conversions,
  metrics.clicks,
  metrics.cost_micros
FROM search_term_view
WHERE segments.date DURING LAST_30_DAYS
  AND metrics.conversions > 0
ORDER BY metrics.conversions DESC
LIMIT 200

Retrieval ladder — if the search-term query returns no rows, follow the shared retrieval ladder in data/search-term-retrieval.md. In pmax-fallback mode, use rows for buyer-language extraction only (language signal, not per-term performance). In limited mode, rely on existing asset performance data for copy direction.

Supplementary: RSA ad-level data:

SELECT
  campaign.name,
  ad_group.name,
  ad_group_ad.ad.responsive_search_ad.headlines,
  ad_group_ad.ad.responsive_search_ad.descriptions,
  ad_group_ad.ad.final_urls,
  metrics.impressions,
  metrics.clicks,
  metrics.conversions
FROM ad_group_ad
WHERE ad_group_ad.ad.type = 'RESPONSIVE_SEARCH_AD'
  AND campaign.status = 'ENABLED'
  AND segments.date DURING LAST_30_DAYS

See data/gaql-recipes.md for additional queries.

Export Mode (no MCP)

Ask the user for:

• RSA asset report (headlines, descriptions, performance labels)
• Or RSA preview from the ads tab
• Search terms report (for buyer language extraction)

Process

1. Announce mode (connected/export).
2. Identify the target query cluster or intent bucket.
3. For search-term buyer language, run the shared retrieval ladder (data/search-term-retrieval.md). In pmax-fallback, use rows for language extraction only. In limited, rely on asset performance data.
4. Extract buyer language and repeated modifiers from converting search terms when available, or from PMax query rows when only language visibility is available.
5. Review current RSA assets: what's BEST, GOOD, LOW, UNRATED?
6. Determine the core promise and LP fit.
7. Recommend RSA components: headline themes, description angles, message hierarchy.
8. Save outputs to workspace memory.

Draft Output

RSA Refresh Draft

Trigger: Analysis shows (a) LOW-performing assets that could be replaced, or (b) buyer language patterns not represented in current ads.

Create using drafts/templates/rsa-draft.md:

• Write to workspace/ads/drafts/YYYY-MM-DD-[account-slug]-rsa-refresh.md
• Include specific headlines (≤30 chars) and descriptions (≤90 chars)
• Note pin recommendations (use sparingly)
• Note which assets to consider removing and why
• Source each new asset to the buyer language it came from
• Update workspace/ads/drafts/_index.md

Always update workspace memory:

• workspace/ads/assets.md — current asset inventory and new recommendations
• workspace/ads/winners.md — high-performing query language
• workspace/ads/queries.md — buyer language patterns

Rules

• Do not write generic ads for mixed intent.
• Use real query language where possible.
• Keep LP alignment explicit.
• If the intent bucket is weak or noisy, say so before generating ads.