Google Ads Audit

This is the broad review. Not just a checklist. A synthesis of what the account is learning, where it is leaking, and what should change.

Read first: - google-ads/references/operator-thesis.md - google-ads/references/intent-map.md - google-ads/references/query-patterns.md - google-ads/references/negatives-playbook.md - google-ads/references/tracking-playbook.md - google-ads/references/structure-playbook.md - google-ads/references/rsa-playbook.md - google-ads/references/budget-playbook.md - google-ads/references/benchmarks.md - google-ads/references/deliverable-templates.md

Read workspace if available: - All workspace/ads/ files except assets.md (unless creative output is needed) - Including workspace/ads/drafts/_index.md — check for existing pending drafts

Data Acquisition

Connected Mode (MCP available)

The audit pulls data across all skill areas. Run queries in this order (each feeds into the next):

1. Account overview:

SELECT
  customer.id,
  customer.descriptive_name,
  customer.currency_code,
  customer.time_zone,
  customer.auto_tagging_enabled
FROM customer

2. Tracking health (check trust first):

SELECT
  conversion_action.name,
  conversion_action.type,
  conversion_action.category,
  conversion_action.counting_type,
  conversion_action.include_in_conversions_metric,
  metrics.conversions,
  metrics.all_conversions
FROM conversion_action
WHERE segments.date DURING LAST_30_DAYS
  AND conversion_action.status = 'ENABLED'
ORDER BY metrics.conversions DESC

3. Campaign performance and structure:

SELECT
  campaign.name,
  campaign.status,
  campaign.advertising_channel_type,
  campaign.bidding_strategy_type,
  campaign_budget.amount_micros,
  metrics.impressions,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions,
  metrics.conversions_value,
  metrics.cost_per_conversion,
  metrics.search_impression_share,
  metrics.search_budget_lost_impression_share
FROM campaign
WHERE campaign.status = 'ENABLED'
  AND segments.date DURING LAST_30_DAYS
ORDER BY metrics.cost_micros DESC

4. Search terms (top 500 by spend):

SELECT
  search_term_view.search_term,
  campaign.name,
  ad_group.name,
  metrics.impressions,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions,
  metrics.conversions_value
FROM search_term_view
WHERE segments.date DURING LAST_30_DAYS
ORDER BY metrics.cost_micros DESC
LIMIT 500

Retrieval ladder — if the search-term query returns no rows, follow the shared retrieval ladder in data/search-term-retrieval.md. Report the resulting retrieval_mode in the audit header. Mark search-term sections as "PMax visibility-limited" when operating in pmax-fallback mode. In limited mode, note the gap and request a UI export.

5. Ad group structure:

SELECT
  campaign.name,
  ad_group.name,
  ad_group.status,
  metrics.impressions,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions
FROM ad_group
WHERE campaign.status = 'ENABLED'
  AND ad_group.status = 'ENABLED'
  AND segments.date DURING LAST_30_DAYS
ORDER BY campaign.name

6. Keyword view (targeted keywords — cross-reference with search terms):

SELECT
  campaign.name,
  ad_group.name,
  ad_group_criterion.keyword.text,
  ad_group_criterion.keyword.match_type,
  ad_group_criterion.status,
  metrics.impressions,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions
FROM keyword_view
WHERE campaign.status = 'ENABLED'
  AND ad_group.status = 'ENABLED'
  AND segments.date DURING LAST_30_DAYS
ORDER BY metrics.cost_micros DESC
LIMIT 200

7. Existing negatives:

SELECT
  campaign.name,
  campaign_criterion.keyword.text,
  campaign_criterion.keyword.match_type
FROM campaign_criterion
WHERE campaign_criterion.negative = TRUE
  AND campaign_criterion.type = 'KEYWORD'

8. RSA assets (if time permits):

SELECT
  asset.text_asset.text,
  asset.type,
  ad_group_ad_asset_view.performance_label,
  ad_group_ad_asset_view.field_type,
  campaign.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

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

Date Range Fallback

All date-ranged queries above default to LAST_30_DAYS. If results are empty or near-zero (<$5 total spend):

1. Re-run with DURING LAST_90_DAYS
2. If still empty, re-run without a date filter (all-time)
3. Always state the date range used in the output header
4. If fallback was needed, note: "Account dormant — using [range] for historical context"

This prevents reporting "no data" on paused or seasonal accounts that have useful history.

Export Mode (no MCP)

Ask the user for as much as they can provide: - Campaign overview (last 30 days) - Search Terms report (last 30 days, top 200+) - Conversion actions list - Current negative keyword list - RSA asset report (optional) - Any notes on tracking, recent changes, or known issues

The audit works with partial data — it will note confidence limitations.

Audit Order

1. Can we trust the account?

If tracking is weak, say so immediately. Assign tracking confidence level (High/Medium/Low/Broken). Note which downstream conclusions are conditional on tracking trust.

2. What is the account learning about buyer intent?

Summarize the query and intent reality. Build or update the Intent Map. Note where intent classes are well-routed vs. poorly routed.

3. Where is waste leaking spend?

Identify junk, ambiguity, and routing failures. Quantify the waste. Separate "waste from bad queries" from "waste from bad structure."

4. What should be negative vs. isolated vs. already wrongly negated?

Distinguish exclusion from structure problems. Default to isolation when in doubt. Also review existing negatives for harm: Check if any current negatives are blocking traffic that matches the business's services. Smart campaign auto-negatives and inherited negatives are common culprits. If harmful negatives are found, include a "Negatives to Remove" section in the negative draft.

5. What structure changes matter most?

Focus on where meaning is being mixed. Prioritize by spend impact, not by cosmetic neatness.

6. Are RSAs aligned with buyer language?

Check whether ads reflect real intent and LP promise. Note specific gaps between search language and ad copy.

7. What deserves more budget, less budget, or no touch yet?

Make support/protect/reduce/hold decisions. Gate budget scaling behind tracking confidence.

8. Are landing pages converting (or do they look like they should be)?

When conversion rates look suspiciously low, run a quick Fork A check (tracking) before blaming the landing page. If clicks are high but conversions are near-zero, the tag may not be firing — not the page's fault. For full landing page → conversion path diagnosis, delegate to /google-ads landing-review. See skills/google-ads-landing-review/SKILL.md and google-ads/references/landing-page-playbook.md.

9. Is PMax helping or contaminating signal?

Be explicit about inference vs. direct evidence. Check for branded/high-intent cannibalization.

Draft Output

Prioritized Draft Batch

The audit produces multiple drafts organized by priority. Not everything needs fixing at once.

Priority tiers: 1. Fix first — tracking problems that gate everything else 2. Quick wins — clear negatives that stop obvious waste 3. Strategic moves — structure changes, budget reallocation 4. Refinement — RSA refreshes, minor optimizations

For each tier with actionable findings, create a draft: - Tracking fixes → drafts/templates/tracking-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-tracking-fix.md - Negatives → drafts/templates/negative-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-negatives.md - Structure → drafts/templates/structure-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-structure.md - Budget → drafts/templates/budget-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-budget-realloc.md - RSAs → drafts/templates/rsa-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-rsa-refresh.md - Landing page → drafts/templates/landing-review-draft.md → workspace/ads/drafts/YYYY-MM-DD-[account-slug]-landing-review.md

Cross-reference dependencies between drafts. E.g., "Apply tracking fix before budget scaling."

Update workspace/ads/drafts/_index.md with all new drafts. If 2 or more drafts are created in the same audit run, also write workspace/ads/drafts/_batch-YYYY-MM-DD-[account-slug].md with: - Account, date range, and tracking confidence - Drafts created in that run grouped by priority tier - Dependency chain and recommended review order - Quick-apply candidates

Do not list the _batch-*.md packet in _index.md. It is an audit packet, not a draft.

Always update workspace memory:

• workspace/ads/intent-map.md — build or update from search terms
• workspace/ads/findings.md — audit findings log
• workspace/ads/negatives.md — recommended negatives
• workspace/ads/learnings.md — what we learned
• workspace/ads/account.md — account context if this is first audit

Output Shape

1. Account Status block — account name, CID, status (active/suspended/dormant), date range used, tracking confidence, mode
2. Executive summary — 3-5 sentences: biggest issues, tracking trust, top priority
3. Tracking confidence detail (High/Medium/Low/Broken)
4. Intent landscape (brief Intent Map)
5. Waste analysis (quantified)
6. Structure assessment
7. RSA alignment check
8. Budget posture
9. PMax assessment
10. Prioritized action plan — organized by tier
11. Drafts created (list with paths, summaries, and dependencies)
12. Confidence assessment
13. Memory updates

Rules

• Tracking trust comes first. Everything else is conditional.
• Search behavior beats generic advice.
• Distinguish symptoms from root causes.
• Do not bury the highest-leverage decision.
• Use benchmarks as context, not as authority.
• The audit is the highest-value skill in the system. It should produce the clearest, most actionable output. Every finding should either become a draft or be explicitly marked as "no action needed."
• Prioritize ruthlessly. An audit that recommends 30 equally-weighted changes is useless. An audit that identifies the 3 things that matter most is gold.