9 Claude Workflows That Find Wasted Google Ads Spend (2027)

Connect Google Ads to Claude through Otto's hosted Claude MCP connector or Google's read-only Python google-ads-mcp server to run nine live GAQL and cross-channel workflows that isolate wasted spend across search terms, AI Max expansions, Performance Max Final URL expansion, Display and YouTube placements, Search Console cannibalization, GA4 conversion gaps, and slow mobile landing pages. Querying filtered metrics via Otto MCP cuts Claude token usage by 80–90% compared to 3,000-row CSV uploads (40,000 to 80,000 tokens) across 227 tools and 15 integrations.

The nine workflows below cover four in-platform Google Ads API diagnostic views, three cross-channel checks connecting Search Console, GA4, and PageSpeed Insights, and two pacing and fix workflows. If you only run one workflow today, run Workflow 1 to triage zero-conversion search terms and pull negative keyword candidates directly from search_term_view.

Five copy-paste prompts let you inspect search queries, automated match expansions, landing-page routing, organic search cannibalization, and destination page speed without leaving your Claude conversation.

Starter prompt 1: Zero-conversion search-term waste

Query Google Ads for the last 30 days using search_term_view. Return the top 25 search terms ordered by metrics.cost_micros descending where metrics.conversions = 0 and metrics.clicks > 5. List the search term, campaign name, clicks, cost, and average CPC. Format the results as a Markdown table, then output a clean list of exact-match negative keyword candidates ready for campaign-level review.

Starter prompt 2: AI Max keywordless and broad-match expansion audit

Run a GAQL query against search_term_view for the last 30 days filtering on segments.search_term_match_source IN ('AI_MAX_KEYWORDLESS', 'AI_MAX_BROAD_MATCH'). Select search_term_view.search_term, segments.search_term_match_source, campaign.name, metrics.cost_micros, metrics.clicks, and metrics.conversions. Group the queries by match source, flag any terms spending budget with zero conversions, and identify whether the search intent aligns with the target campaign.

Starter prompt 3: Performance Max Final URL expansion leak check

Query final_url_expansion_asset_view and campaign asset automation settings for all Performance Max campaigns over the last 30 days. List each dynamically selected landing page, the generated headline and description text assets, clicks, cost, and conversions. Identify destination URLs that point to checkout, blog, or support pages, and output WEBPAGE campaign criteria exclusion rules to block them.

Starter prompt 4: Google Ads vs. Google Search Console cannibalization check

Compare paid search queries from Google Ads search_term_view with organic search queries from Google Search Console over the last 28 days. Identify queries where the organic ranking is in position 1 to 3 with an organic CTR above 15%, while Google Ads also spent budget on the exact query. Calculate the total paid spend on these organic terms and categorize which queries represent brand defense versus potential paid cannibalization.

Starter prompt 5: Destination URL Core Web Vitals and GA4 conversion check

Identify the top five Google Ads landing pages by spend over the last 30 days that have a conversion rate below 1%. For each landing page URL, run psi_analyze_url to retrieve mobile Lighthouse scores, Largest Contentful Paint, and Cumulative Layout Shift. Cross-check GA4 key events for each URL and flag pages where poor mobile performance correlates with dropped paid conversions.

Growth marketers and service businesses lose budget when search queries drift, dynamic landing pages misroute traffic, or landing pages convert poorly after the click. As Head of Growth & Marketing at IslaIntel, Guillermo Marti leads paid ads, SEO, and AI marketing workflows across Meta, Google, and Claude MCP at Otto, where Otto Webmaster operates as an AI digital webmaster and marketing teammate; this guide represents Otto's Funnel-track workflows bylined to Marti that feed /blogs/funnel.json.

Practitioners traditionally audit waste by downloading 3,000-row CSV reports from Google Ads, reformatting columns in spreadsheets, and uploading files into AI chats. Uploading a raw 3,000-row Ads Manager CSV into Claude can consume 40,000 to 80,000 tokens in a single turn, whereas querying aggregated metrics through Otto MCP cuts token usage by 80–90%. Google Ads reporting queries construct requests via Google Ads Query Language (GAQL) selecting specific attributes, segments, and metrics from target resources through GoogleAdsService.SearchStream or GoogleAdsService.Search, as documented in the Google Ads API reporting overview.

The table below contrasts manual CSV analysis, Google's open-source Python MCP server, and Claude connected to Otto MCP.

According to the Google Ads developer toolkit MCP server documentation, Google's open-source Python MCP implementation is strictly read-only, exposes list_accessible_customers, search, and get_resource_metadata, and cannot modify bids, pause campaigns, or create new assets. Otto Webmaster connects over HTTP at https://otto.islaintel.com/api/mcp using a single organization-scoped bearer token prefixed otto_mcp__. Users connect accounts inside Otto via OAuth, keeping secrets on the server so keys are never pasted into Claude chats. Documented AI clients supported by Otto MCP include Claude, ChatGPT, Cursor, Claude Code, OpenClaw, and Hermes, while Otto core is also accessible via WhatsApp, web app, and email.

Connect Claude to Otto MCP

Google Ads API diagnostic resources allow growth teams to isolate automated query matching and placement leaks directly in conversation.

1. Search Terms and Negative Keyword Triage (search_term_view)

Search-term triage isolates search queries that consumed budget without generating a conversion.

• What it catches: Broad-match query drift, irrelevant queries, and search terms that accumulate spend across ad groups without conversions.
• Resource used: search_term_view queried via GAQL attributes, segments, and metrics via GoogleAdsService.SearchStream or GoogleAdsService.Search as documented in the Google Ads API reporting overview.
• The catch: Google Ads is strictly read-only in Otto MCP with no spend or configuration changes, matching Google's open-source Python google-ads-mcp server which cannot modify bids, pause campaigns, or create new assets.

Claude prompt:

Use the Google Ads search tool to run this GAQL query:
SELECT
  search_term_view.search_term,
  campaign.name,
  ad_group.name,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions,
  metrics.average_cpc
FROM search_term_view
WHERE segments.date DURING LAST_30_DAYS
  AND metrics.cost_micros > 20000000
  AND metrics.conversions = 0
ORDER BY metrics.cost_micros DESC
LIMIT 50

Analyze the returned queries. Group them into three themes: completely irrelevant search intent, informational queries that require blog content rather than paid traffic, and competitor brand queries. Provide a clean list of exact-match negative keywords ready to paste into campaign negative keyword lists.

2. AI Max for Search Expansion Audit (AI_MAX_KEYWORDLESS and AI_MAX_BROAD_MATCH)

Google Ads AI Max for Search introduces automated matching sources that expand targeting beyond regular match types. Advertisers learn locking Google Ads AI Max landing pages and controlling Final URL expansion to avoid drift.

• What it catches: Unintended keywordless traffic pulled from website content and aggressive broad-match expansions.
• Resource used: search_term_view filtered by segments.search_term_match_source, where the Google Ads API AI Max reporting documentation defines AI_MAX_KEYWORDLESS for landing-page matches without keywords and AI_MAX_BROAD_MATCH for AI-expanded keywords.
• The catch: Google Ads is read-only in Otto MCP with no spend or configuration changes, so Claude drafts negative keyword lists while bid and campaign updates stay paused for your manual action.

Claude prompt:

Use GAQL to query search_term_view for the last 30 days:
SELECT
  search_term_view.search_term,
  segments.search_term_match_source,
  campaign.name,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions,
  metrics.cost_per_conversion
FROM search_term_view
WHERE segments.date DURING LAST_30_DAYS
  AND segments.search_term_match_source IN ('AI_MAX_KEYWORDLESS', 'AI_MAX_BROAD_MATCH')
ORDER BY metrics.cost_micros DESC
LIMIT 50

Break down total spend, clicks, and conversions by match source. Highlight queries where AI_MAX_KEYWORDLESS matched irrelevant website subpages. Suggest negative keywords and note if landing-page exclusions are required.

3. Performance Max Final URL Expansion and Dynamic Asset Audit (final_url_expansion_asset_view)

Performance Max automatically expands destination URLs unless campaign settings restrict traffic.

• What it catches: Ad spend diverted away from primary landing pages toward dynamically selected landing pages.
• Resource used: final_url_expansion_asset_view, where the Google Ads API Performance Max optimizations guide notes Final URL expansion is active by default (FINAL_URL_EXPANSION_TEXT_ASSET_AUTOMATION), generates dynamic text assets, and supports WEBPAGE campaign criteria exclusions.
• The catch: Excluding non-converting dynamic pages requires configuring WEBPAGE campaign criteria in Google Ads, which remains strictly read-only in Otto MCP.

Claude prompt:

Query final_url_expansion_asset_view in Google Ads for the last 30 days:
SELECT
  campaign.name,
  final_url_expansion_asset_view.final_url,
  final_url_expansion_asset_view.asset,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions
FROM final_url_expansion_asset_view
WHERE segments.date DURING LAST_30_DAYS
ORDER BY metrics.cost_micros DESC
LIMIT 50

List every landing page URL selected by Performance Max URL expansion. Flag non-converting pages such as terms of service, utility pages, or blog posts. Draft the exact WEBPAGE exclusion criteria URLs needed to prevent Performance Max from directing paid traffic to those paths.

4. Display, YouTube, and App Placement Leak Audit (group_placement_view)

Automatic targeting across the Google Display Network and YouTube frequently serves impressions inside mobile games and low-quality websites.

• What it catches: Accidental clicks from mobile utility apps, low-intent mobile gaming apps, and irrelevant YouTube channels.
• Resource used: group_placement_view, which shows where ads served across websites, YouTube channels, and apps whether explicitly targeted or discovered automatically as detailed in the Google Ads API placement views documentation.
• The catch: Google Ads reporting tools in Otto MCP are strictly read-only, meaning Claude flags low-quality placement domains while placement exclusions require direct campaign management.

Claude prompt:

Query group_placement_view for Display, Video, and Performance Max campaigns over the last 30 days:
SELECT
  group_placement_view.placement,
  group_placement_view.placement_type,
  group_placement_view.target_url,
  campaign.name,
  metrics.clicks,
  metrics.cost_micros,
  metrics.conversions
FROM group_placement_view
WHERE segments.date DURING LAST_30_DAYS
  AND metrics.cost_micros > 10000000
ORDER BY metrics.cost_micros DESC
LIMIT 50

Identify placements matching mobile app packages (such as 'mobileapp::'), kid-focused YouTube channels, or MFA (made-for-advertising) content domains. Calculate the wasted budget across these placements and generate placement exclusion recommendations.

Isolated ad platform reporting hides root causes. Claude running Otto MCP can cross-reference paid search data with organic search positions, web analytics, and technical page speed in the same conversation.

5. Paid vs. Organic Search Cannibalization Check

Paying for clicks on queries where your site already ranks organically in position one can waste money that could fund incremental terms.

• What it catches: Heavy ad spend on branded or core commercial queries where organic listings already capture demand.
• Tools used: Google Ads search_term_view and Google Search Console search analytics tools via Otto MCP.
• The catch: Google Ads and Google Search Console are read-only in Otto MCP with no spend or configuration changes; Otto structures websites, schema, listings, and reviews without promising rankings or citations, while automated SEO reports track keyword progress over time.

Claude prompt:

Step 1: Retrieve the top 30 Google Ads search terms by spend over the last 30 days from search_term_view.
Step 2: Query Google Search Console for the same date range to fetch organic clicks, impressions, CTR, and average position for those exact queries.
Step 3: Build a comparison table showing Query, Paid Spend, Paid Clicks, Organic Position, and Organic CTR.
Step 4: Flag queries where organic position is between 1.0 and 2.0 with an organic CTR above 20%. Advise whether to lower bids, apply negative keywords, or maintain ad coverage for defensive positioning.

6. Google Ads Conversions vs. GA4 Landing-Page Key-Event Reconciliation

Google Ads conversion tags can report modeled numbers that diverge from GA4 key events. Marketers maintain a GA4 and paid media landing-page dashboard workflow to spot discrepancy leaks.

• What it catches: Tracking failures, session drops between ad clicks and page loads, and campaigns claiming credit for unverified conversions.
• Tools used: Google Ads reporting tools and GA4 run report tools in Otto MCP.
• The catch: Google Ads and GA4 are read-only in Otto MCP with no spend or configuration changes, allowing Otto to perform weekly ad checks and reporting while strategic decisions remain with the owner.

Claude prompt:

Query Google Ads for campaign-level clicks, cost, and conversions over the last 30 days.
Next, query GA4 for the same period filtering by sessionSource / sessionMedium = 'google / cpc'. Retrieve GA4 sessions, engaged sessions, and key events grouped by campaign and landing page path.
Compare the two datasets:
1. Identify campaigns where Google Ads reported conversions exceed GA4 key events by more than 25%.
2. Calculate the drop-off percentage between Google Ads clicks and GA4 recorded sessions.
3. Highlight landing pages where session engagement drops below 40%, indicating tracking tag failure or immediate visitor bounce.

7. Mobile Landing-Page Core Web Vitals Audit (psi_analyze_url)

Slow landing pages reduce conversion rates, inflating your cost per acquisition. Review how Core Web Vitals drift causes paid conversion leaks before adjusting ad bids.

• What it catches: Landing pages with Largest Contentful Paint delays, layout shifts, or uncompressed media that cause mobile visitors to bounce before converting.
• Tools used: Otto MCP's built-in psi_analyze_url tool, which inspects Lighthouse scores and Core Web Vitals on any URL using Otto's platform key without requiring a user Google API key.
• The catch: Otto MCP's psi_analyze_url runs in read-only mode with no spend or configuration changes; fixing a slow landing page requires staging a WordPress draft or a GitHub pull request with a preview for human approval.

Claude prompt:

Run psi_analyze_url on our top three Google Ads landing page URLs:
1. [Insert URL 1]
2. [Insert URL 2]
3. [Insert URL 3]

Analyze mobile performance metrics:
- Overall Lighthouse Performance score
- Largest Contentful Paint (LCP)
- Cumulative Layout Shift (CLS)
- Total Blocking Time (TBT)

Correlate slow load metrics with paid conversion drop-offs. Provide specific frontend optimization recommendations (such as image compression, script deferral, or DOM reduction) that we can stage on our website.

Cross-channel analysis and staging tools turn diagnostic insights into concrete optimizations without risking production stability.

8. Cross-Channel Budget and CPA Drift Audit

Evaluating Google Ads alongside other paid channels prevents overspending on underperforming networks. Marketers also reference 10 Claude workflows for Meta Ads in 2027 when coordinating campaigns.

• What it catches: Channel-level CPA inflation where Google Ads outspends Meta or LinkedIn Ads despite delivering higher acquisition costs.
• Tools used: Google Ads, Meta Ads, and LinkedIn Ads reporting tools in Otto MCP.
• The catch: Google Ads is read-only in Otto MCP, Trial ($1 for 3 days) and Pro ($199/mo, or $99/mo billed annually) provide read-only ads access, and write access to Meta and LinkedIn Ads requires the Master plan ($3,500/mo, or $3,000/mo billed annually) with reviewer or admin approval.

Claude prompt:

Pull campaign spend, clicks, conversions, and calculate CPA for Google Ads, Meta Ads, and LinkedIn Ads over the last 14 days.
Format a unified cross-channel performance table:
- Channel
- Campaign Name
- Total Spend
- Conversions
- Effective CPA

Highlight campaigns where effective CPA exceeds target thresholds by more than 30%. Identify which channel delivers the most efficient conversions for top-of-funnel acquisition versus bottom-of-funnel conversion. Suggest budget reallocations across channels.

9. Turning Wasted-Spend Diagnostics Into Staged Landing-Page Fixes

Unlike ad-only tools, Otto also fixes the landing page after diagnosing the ad by opening a GitHub pull request or WordPress draft in the same workflow.

• What it catches: Conversion loss caused by mismatched landing-page headlines, missing call-to-action buttons, or weak offer alignment.
• Tools used: Otto MCP's WordPress integration (29 tools: 15 read tools and 14 write tools) or GitHub integration with Vercel deployment preview orchestration.
• The catch: Every site update waits on approval: GitHub edits open as pull requests with preview orchestration, WordPress content defaults to draft across 29 tools, and checkout, contact, and legal pages always require human review in Otto core.

Claude prompt:

Based on the high-cost, low-converting search terms identified in Workflow 1, let's update the target WordPress landing page (Page ID: 1042) to improve relevance:
1. Read the current page content using Otto's WordPress tools.
2. Draft an updated hero headline, subheadline, and proof bullet points incorporating high-intent search queries.
3. Stage the update as a WordPress draft so no changes go live immediately.
4. Output a summary of the edits and the draft preview link for owner approval.

How to choose which workflow to run first

Account priority dictates which workflow to start with:

• If your search ad spend is spiking: Run Workflow 1 (search_term_view) to identify non-converting queries and build a negative keyword list.
• If you recently turned on automated matching: Run Workflow 2 (AI Max audit) and Workflow 3 (Performance Max Final URL expansion audit) to block unwanted matches.
• If your click volume is steady but conversions dropped: Run Workflow 6 (GA4 reconciliation) and Workflow 7 (Core Web Vitals audit) to isolate landing-page bottlenecks.

Marketers need certainty that automated AI assistants will not execute unapproved changes or drain advertising budgets.

Read-only guardrails and approval gates

Otto enforces strict permission boundaries across connected platforms:

• Strictly read-only platforms: In Otto MCP, Google Analytics 4, Google Search Console, Google Ads, Google Business Profile, YouTube, Google Sheets, Shopify, and PageSpeed Insights are read-only with no spend or configuration changes. This matches Google's open-source Python google-ads-mcp server, which is strictly read-only and cannot modify bids, pause campaigns, or create new assets.
• Approval-gated ad platforms: In Otto MCP, LinkedIn Ads campaigns and ad sets are created as DRAFT by default, activating or pausing requires confirm=true plus a reviewer or admin, and $0 budgets are rejected. Meta Ads write operations also require reviewer or admin access.
• Social media publishing: Unlike WordPress (where new posts and pages default to draft), publishing to Facebook Pages and Instagram through Otto MCP goes live immediately and requires reviewer or admin permissions.
• Website staging: For website updates, every change on GitHub opens as a pull request with a preview, and on WordPress new content stays a draft until the owner publishes it. Otto integrates with Vercel to link repositories and support preview orchestration. Checkout, contact, and legal pages always require human approval in Otto core, and otto_get_change_log records recent Webby deploys and site-edit runs.
• Direct mobile approvals: On June 2, 2026 (v1.1.0), Otto shipped Enhanced WhatsApp Integration allowing users to approve or reject content changes directly from WhatsApp without logging into the web dashboard.

Otto reads the accounts a business connects, proposes what to change, and ships changes only after the owner approves: new ads stay paused, WordPress posts stay drafts, and site edits wait on a preview. Otto is not a website builder; he works on the site a business already has on its existing platform, drafts the change, and waits for approval. Otto performs the weekly execution that marketing retainers typically cover—site updates, SEO maintenance, ad checks, and reporting—while strategy and final approval stay with the business owner.

Explore how the Otto AI Paid Ads Manager operates alongside website management.

How Otto compares to ad-only tools

Many specialized tools audit paid advertising. The following comparison highlights structural differences between Otto and alternatives:

Opteo monitors Google Ads accounts and suggests bid, negative keyword, and search-term fixes, but covers Google Ads only with no Meta Ads, SEO, website management, or Claude MCP server. Synter offers paid media connectors across 40+ ad networks for Claude, ChatGPT, and Cursor users, but does not manage websites or technical SEO deploys. Ryze AI focuses primarily on paid ads and SEO/GEO workflows, whereas Otto also manages live website edits (GitHub pull requests, WordPress pages/posts, and Google Business Profile) and organic social publishing. Madgicx combines Meta Ads creative analytics with rule-based bid and budget automation for e-commerce stores.

Review our evaluation of the 7 best Ryze AI alternatives for paid ads and SEO for detailed comparisons.

Check our transparent Otto Trial, Pro, and Master plans to review platform capabilities.

Pricing structure

Otto MCP uses the same Trial, Pro, and Master plans as the core Otto website product rather than a separate MCP subscription price:

• Trial: $1 for 3 days (a trial of Pro: website updates and SEO, plus read-only access to ads and socials). A homepage redesign is included with the $1 Otto trial, built from the business's own photos and words and approved by the owner before going live.
• Pro: $199/mo, or $99/mo billed annually. Includes website updates, SEO, and read-only access to ads and socials. Users can cancel anytime.
• Master: $3,500/mo, or $3,000/mo billed annually. Adds approval-gated write access to ads and socials with human approval on every change. Users can cancel anytime.

Honest limitations

Otto maintains clear operational limits:

• Ad platform write boundaries: In Otto MCP, Google Ads is strictly read-only with no spend or configuration changes. Write access to ads and socials requires the Master plan.
• Agency white-labeling: White-label agency portals are limited.
• Terminal CLI status: The Otto CLI for terminal agents is coming soon rather than shipping today; MCP is the supported path today. Otto is not Otto The Agent and not OttoKit.
• Unsupported platforms: Otto does not support unannounced third-party integrations such as Wix, Webflow, or TikTok in the live MCP catalog.
• No ranking or citation promises: Otto does not promise Google Map Pack placement, search rankings, or ChatGPT mentions; it structures websites, schema, listings, and reviews so search engines and AI assistants can read and understand the business.

Frequently asked questions

How does Claude connect to Google Ads through Otto MCP?

OAuth connects your Google Ads account inside the Otto dashboard. Otto provisions an HTTP endpoint (https://otto.islaintel.com/api/mcp) using an organization-scoped token prefixed otto_mcp__. Claude communicates over HTTP, executing GAQL queries without exposing sensitive OAuth secrets.

Can Claude accidentally increase my Google Ads budget or pause campaigns?

Google Ads inside Otto MCP is strictly read-only with no spend or configuration changes. Claude cannot adjust budgets, pause campaigns, modify bids, or create ad assets in Google Ads.

Why use Otto MCP instead of uploading a Google Ads CSV into Claude?

Uploading a raw 3,000-row Ads Manager CSV consumes 40,000 to 80,000 tokens in a single turn, whereas querying aggregated metrics through Otto MCP cuts token usage by 80–90%. Otto MCP also connects 227 tools across 15 integrations in one conversation.

What happens after Claude finds a wasted-spend leak on our landing page?

Otto stages a fix on your existing platform without publishing live edits. For WordPress sites, Otto drafts new pages or updates existing posts as drafts across 29 tools. For GitHub repositories, Otto opens a pull request with a preview deployment for your review and approval.

Connect Claude to Otto MCP