Audit Your Google Ads Account in 5 Claude Prompts (Real Walkthrough)
A real Google Ads audit takes a senior PPC analyst 4-8 hours and costs $500-2,500 if you outsource it. With Claude + MCP, the same audit takes about 20 minutes and costs roughly $1 in API tokens. Below: the five prompts that cover 90% of what a manual audit surfaces, the real outputs we saw running them on a $14K/month B2B SaaS account, and the actions we took from each.
Performance marketing lead with 8+ years running paid media for DTC, SaaS, and high-ticket coaching brands.
Table of Contents
Before you run these — the setup
Three prerequisites:
- Claude Pro, Claude Code, or another MCP-compatible AI client
- An MCP server connected to your Google Ads account (we use 1ClickReport — $25/mo Pro tier covers all 5 prompts; $99/mo Premium adds the ability to execute changes from the same conversation)
- OAuth-connected access to the Google Ads account you're auditing
Setup time: ~5 minutes for first-time connection. After that, the audit is purely prompt-driven.
Prompt 1: Top wasted-spend keywords
Pull Google Ads data for the last 30 days. List keywords where:
- Spend > $50
- Conversions = 0
- Impressions > 200
Show: keyword, match type, ad group, campaign, spend, CTR, CPC.
Sort by spend descending. Limit 20 results.
For each, recommend: pause, add as negative keyword, or investigate further.What we saw on the test account: 14 keywords spent a combined $1,820 with zero conversions. The top one alone burned $340 on broad match for a term that turned out to be a generic industry phrase, not buyer intent. We paused 11 immediately and converted 3 to negative keywords (those were attracting wrong-intent traffic).
Recovery: ~$1,500/month in wasted spend identified in 30 seconds.
Prompt 2: Search-term cleanup
From Google Ads last 30 days, pull all search terms with:
- Impressions > 100
- CTR < 0.5%
Show: search term, the keyword that triggered it, impressions, clicks,
spend. Sort by impressions descending. Limit 30.
For each, recommend: add as negative keyword, restructure the keyword
that triggered it, or accept (if the search term is legitimate but low-quality).What we saw: 67 search terms with high impression but bad CTR — most because the trigger keyword was too broad. Top recommendation: add 41 as negative keywords across the account. The remaining 26 were legitimate but indicated the broad-match keyword needed splitting into exact-match variants.
Impact: cleaned up the impression pool, freed budget for higher-intent searches.
Prompt 3: Quality Score issues
List all Google Ads keywords from active ad groups with Quality Score <= 5.
Show: keyword, ad group, campaign, current QS, expected CTR component,
ad relevance component, landing page experience component, last 30-day
spend, CPC.
Sort by 30-day spend descending. Limit 25.
For each, identify the lowest-scoring QS component and recommend a
specific fix (rewrite ad copy, improve landing page relevance, restructure
ad group).What we saw: 18 keywords with QS ≤ 5 collectively burning $4,200/month at inflated CPCs. The breakdown:
- 11 had "below average" ad relevance — fix was tighter ad copy aligned to the keyword
- 5 had "below average" landing page experience — landing page didn't mention the keyword's intent
- 2 had "below average" expected CTR — historical performance issue, recommended pausing and re-creating with different match type
Estimated CPC reduction from fixes: 15-25% on the affected keywords. Net impact: more clicks for the same spend.
Prompt 4: Bidding strategy fit
For each Google Ads campaign in my account, show:
- Current bidding strategy
- Last 30 days: spend, conversions, conversion value, ROAS
- Conversions per week average (to assess whether strategy has enough data)
Recommend: keep, switch to Maximize Conversions, switch to Target ROAS,
switch to Manual CPC, or split into multiple campaigns. Explain reasoning
per campaign.What we saw: 8 campaigns, 3 of which were on Target ROAS but receiving fewer than 30 conversions/month — well below the threshold for Target ROAS to work well. Recommendation: switch those to Maximize Conversions until they hit 50+ conversions/month, then revisit.
One campaign on Manual CPC had 200+ conversions/month — clear candidate to move to Target ROAS. Implemented two weeks later: ROAS improved 18%.
Prompt 5: Impression share gaps
List Google Ads campaigns where:
- Search lost impression share (budget) > 20% OR
- Search lost impression share (rank) > 30%
Show: campaign, type, lost IS budget, lost IS rank, last 30-day spend,
ROAS. Sort by ROAS descending.
For each, recommend specific action: raise daily budget, raise bid,
improve QS to win more auctions, restructure to be more competitive.What we saw: 4 campaigns leaking impression share. Two were ROAS > 3.5x — clear "we're profitable, give them more budget" decisions. One was ROAS 2.1x and losing rank IS — recommendation was QS improvement first, bid raise second. One was actually fine — losing IS because they'd hit budget cap intentionally; no action needed.
Net action: reallocated $3,000/month from two underperforming campaigns to the two ROAS leaders. Projected incremental revenue: $9,000-12,000/month.
What 20 minutes of Claude replaces
Manual audit equivalents and their typical cost:
| Activity | Manual time | Outsource cost |
|---|---|---|
| Wasted-spend audit (manual via search terms + keyword reports) | 2 hours | $200-400 |
| Search-term cleanup | 1-2 hours | $100-300 |
| Quality Score audit + fix recommendations | 2-3 hours | $300-600 |
| Bidding strategy review | 1 hour | $200-400 |
| Impression share + budget allocation review | 1-2 hours | $200-400 |
| Full equivalent audit | 7-10 hours | $1,000-2,100 |
For 20 minutes of your time + ~$1 in API tokens.
The honest tradeoff: the manual analyst will catch a few things the prompts miss (intuition about specific industries, knowledge of unusual account history). For 95% of accounts, the prompts get you within striking distance and the value of the time saved compounds across every audit you run.
Running this weekly
Most accounts benefit from running prompts #1 and #2 weekly (catch new wasted spend before it accumulates), and prompts #3, #4, #5 monthly (deeper changes that need more data to validate).
Total weekly time investment: 5-10 minutes. Monthly: 20-30 minutes total.
Compounding effect: accounts audited weekly tend to maintain 15-30% better ROAS than accounts audited quarterly, simply because issues get caught and fixed before they snowball.
Frequently Asked Questions
Can I run these without an MCP?
Not effectively. Claude without MCP can guide you through what to look for, but it can't pull the data itself. You'd be doing the data extraction manually and feeding it into Claude — which loses most of the time savings.
Will Claude actually pause keywords for me?
Only with the Premium plan ($99/mo on 1ClickReport) that enables campaign management. The Pro plan ($25/mo) is read-only — Claude can recommend the pause but you'd execute it in Google Ads UI. Most accounts start with Pro and upgrade once the value of execute-in-place is clear.
Does this work for shopping campaigns?
Mostly yes — the prompts above are written for Search campaigns. Shopping campaigns have different metrics (product groups vs keywords) but the same approach works with light prompt adjustments (replace 'keyword' with 'product group').
What about Performance Max campaigns?
Performance Max has limited visibility by design — Google doesn't expose per-asset performance the way Search does. The prompts above work for the campaign-level views but you won't get keyword-level insights from PMax.
How often should I re-run these prompts?
Prompt 1 (wasted spend) and 2 (search terms) weekly. Prompts 3-5 monthly. The wasted-spend prompt is the highest-ROI single prompt in the set — it pays for itself most weeks.
Will Google flag the account for unusual activity if Claude runs these?
No — these prompts pull data via the official Google Ads API through OAuth. From Google's perspective, this is standard authenticated API usage. The only thing that could trigger flags is the Premium plan's campaign-edit actions, which use the same Google Ads API endpoints as Google's own UI.