Skip the install. Get this working in under 2 minutes.
Start a free trial on cloud.anythingmcp.com, add the Google Ads in one click, then point your AI client (Claude, ChatGPT, Copilot or Cursor) at the generated MCP endpoint. No Docker, no git clone, zero engineering experience required.
Summary
Audit a Google Ads account with Claude in minutes: spend, conversion quality, wasted search terms, Quality Score, budgets and tracking, straight from the API.
Try asking
Example prompts for Google Ads
Click any prompt to copy it. Paste into Claude, ChatGPT, Cursor, Gemini, Copilot or OpenClaw to run it against this connector.
Claude is AI and can make mistakes. Please double-check responses.
A Google Ads audit in one conversation
Short answer: connect Google Ads to Claude through AnythingMCP, then ask Claude to audit the account. Claude reads the account settings, the daily trend, campaign totals, conversion actions, search terms, keywords and budgets through read-only tools, and reports what looks wrong with the numbers to back it. Nothing in the account is changed; you decide what to do with the findings.
A manual audit means exporting a dozen reports and cross-checking them. Claude can pull each report when it needs it, for any date range, and follow a lead: if the cost per conversion looks too good, it checks the days and the conversion actions before you see the summary.
What we asked
We ran this on our own account, in claude.ai, with the Google Ads connector installed:
With the Google Ads connector: how did the account do in the last 30 days? Spend, conversions, cost per conversion and the main campaigns. Read only, no changes.
What Claude did
| Step | Tool | Why |
|---|---|---|
| 1 | gads_list_accessible_customers | find the accounts the user can open |
| 2 | gads_get_customer | currency and time zone of each account |
| 3 | gads_campaign_performance | totals per campaign for the 30 days |
| 4 | gads_list_campaigns | which campaigns are enabled or paused, and why |
| 5 | gads_account_daily_metrics | the daily trend behind the totals |
| 6 | gads_conversion_actions | what is actually being counted as a conversion |
Six calls, all reads, well inside the daily quota.
What it found
- Two accounts with the same name. One had a single paused campaign and no activity all year; Claude said so and used the other one.
- Totals. Spend, conversions, clicks, CTR and cost per conversion for the 30 days, with the currency named.
- One campaign did all the work. A Performance Max campaign carried all spend and conversions; the Search campaign was paused.
- Numbers too good to be true. The cost per conversion was very low for a B2B account. In the daily trend Claude found two short bursts with clicks at about one cent each that produced most of the month's conversions, and a two-day gap where the campaign barely ran.
- What the conversions were. Almost every conversion was one sign-up action, tracked as a page visit, with a fixed default value. Claude recommended comparing Google's sign-up count with the real sign-ups in the product for those days before trusting the campaign's optimization, because Performance Max optimizes toward whatever the primary conversion counts.
That last point is the one that matters: the campaign looked excellent on the dashboard, and the audit showed why the number could not be taken at face value.
The full audit checklist
The connector's playbook (gads_playbook) gives Claude the queries for each part. Ask for any of these by name:
| Area | What Claude checks | Main tools |
|---|---|---|
| Overview | totals, trend, top campaigns, CPA and ROAS | gads_account_daily_metrics, gads_campaign_performance |
| Conversion quality | primary actions, counting type, value, volume per action | gads_conversion_actions, gads_run_gaql |
| Wasted spend | search terms and keywords with cost and no conversions | gads_search_terms, gads_keyword_performance |
| Negative keywords | recurring bad terms, existing negatives to avoid duplicates | gads_search_terms, gads_run_gaql |
| Quality Score | low scores and the component that drags them down | gads_keyword_performance |
| Budget and rank | impression share lost to budget vs to rank | gads_budget_pacing |
| Ads | disapprovals, ad strength, ad groups with one ad | gads_ad_performance |
| Targeting | device, location (including spend outside targeted countries), hour and weekday | gads_run_gaql |
| Performance Max | asset groups and search categories | gads_run_gaql |
| Changes | who changed what around a drop | gads_change_history |
Prompts for the next steps
- "Audit my Google Ads account for the last 30 days using the playbook: overview, conversion quality, wasted spend, budgets and ads. Read only."
- "Show spend and conversions by country for the Performance Max campaign, and flag countries we do not target."
- "List search terms with spend and no conversions in the last 90 days and group them into negative keyword themes."
- "Which keywords have Quality Score 5 or lower, and which component is below average?"
- "Compare this month with the same period last year, by campaign."
FAQ
Is it safe to let Claude into the ad account? The connector has no write tools, so Claude cannot change campaigns, bids or budgets. For extra safety, authorize it with a Google Ads user that has the Read only role.
How long does it take? The run above took under a minute of tool calls. A full checklist audit is a few minutes and a few dozen API operations.
Does it work for agencies with many accounts? Yes. With a manager account, Claude lists the client accounts and runs the same checks per account; mind the daily quota of your Cloud project (2,880 operations on Explorer access).
Do I need a developer token? No. Since September 2026 the Google Ads API uses your Google Cloud project's access level instead; see Google Ads API access without a developer token.
Related
- Google Ads MCP server: the overview
- Connect Google Ads to Claude
- Connect Google Ads to ChatGPT
- Google Ads in the marketplace
Was this guide helpful?
