Blog

Best AI Ads Tools Team · May 17, 2026

Launch Google Ads with AI: How We Built a Search Campaign in 1 Hour Using Claude + Supermetrics

Most people spend weeks learning Google Ads before launching their first search campaign. A proper search setup requires keyword research, strategic grouping, negative keyword sculpting, ad copy variation, and precise targeting. Done manually, that is several hours of detailed work where one mistake can derail performance. We did it in one hour.

Not by cutting corners. By using Claude as a campaign manager with Supermetrics’ Model Context Protocol (MCP) integration. Claude handled the keyword research, proposed the keyword grouping, built the entire campaign structure, applied negative keywords, wrote the ads, and verified everything through the API. We handled the strategy, approvals, and the three settings Claude cannot touch.

This is not a theoretical guide. This is the exact workflow we used to launch a real campaign for CryptoJobsList.com. What worked, what broke, and what you should know before you try it yourself.

Recommended AI Ad Tools

Tool	Rating	Price	Best For
Superscale	4.9/5	$49/mo	End-to-end AI ad platform	Try Superscale
Jasper AI	4.7/5	$39/mo	AI marketing copy	Try Jasper AI
Anyword	4.4/5	$49/mo	Predictive AI copy	Try Anyword

The Approach: Claude as Campaign Orchestrator

The sequence looks like this:

Business context (us)
Creative keyword brainstorm (Claude)
Volume validation (Claude + your CSV + live API check)
Audit of existing campaigns (Claude)
Ad group structure design (Claude proposes, we approve)
Bulk execution (Claude via API)
Verification (Claude re-pulls state to confirm)
Tuning pass (Claude suggests, we decide)

The key insight: Claude is not replacing your judgment. Claude is replacing the tedious, error-prone setup work that normally takes hours in the Google Ads UI, keyword research tools, and spreadsheets. One misgrouped keyword or missing negative can waste budget for weeks. Claude gets the structure right the first time.

How We Used Claude in Practice

You drive decisions. Claude executes. That is the repeated pattern.

Example from our build:

Claude flags an issue: “Your existing campaign targets Spain but your landing page is in English and your pricing is in USD. Should we fix the geo targeting?”

We approve: “Yes, switch to US, UK, Canada only.”

Claude executes: API call to update geo targeting.

Claude verifies: Re-pulls campaign state and reports back: “Geo targeting updated. Campaign now targets US, UK, Canada. Impression share for English keywords should improve significantly.”

This pattern happens 20+ times during a campaign build. Claude proposes, you green-light, Claude acts, Claude confirms.

Where you must stay in control:

Budget decisions (Claude suggests, you set the number)
Keyword approvals (Claude generates ideas, you reject the weak ones)
Ad copy tone (Claude writes, you edit for your brand voice)
Landing page selection (Claude cannot see your page quality)

The Setup: Supermetrics MCP + Google Ads

What you need:

A Google Ads account
Supermetrics account (hub.supermetrics.com)
Claude with MCP tool access

Step 1: Connect Google Ads in Supermetrics

Go to https://hub.supermetrics.com/token-management and connect your Google Ads account. Use the data source code “AW” for Google Ads.

Step 2: Enable write access

This is the step everyone misses. Go to https://hub.supermetrics.com/write-settings and explicitly enable write access for Google Ads. Without this, Claude can read your campaigns but cannot create or edit anything. This was our mid-build blocker.

Step 3: Configure workspace folder

Tell Claude where to save progress files. We used /Users/seb/DEV/CryptoJobsList.com. Claude saved a summary file and a to-do list after each major phase. This meant we could stop mid-build, restart later, and Claude would read the progress file to pick up where we left off.

What This Setup Can Do (The Full List)

Once connected, Claude can:

Read everything: campaigns, ad groups, keywords, ads, extensions, performance data
Pull keyword search volumes live from Google
Get keyword ideas based on your landing page or seed terms
Pull historical performance: impressions, clicks, conversions, search terms reports
Create campaigns, ad groups, and ads in bulk
Update and pause existing campaigns, ad groups, and ads
Apply negative keywords at campaign and ad group level
Set bidding strategies: Maximize Clicks, Maximize Conversions, Target CPA, Target ROAS, Manual CPC
Configure geo targeting by country, region, or city
Set language targeting
Configure network settings (Search, Display, Search Partners)
Add extensions: sitelinks, callouts, structured snippets
Run async data queries for arbitrary date ranges

This covers about 85% of what you do in a typical campaign build.

What This Setup Cannot Do (The Limitations)

These are the Google Ads features Claude cannot reach. You must handle these manually in the UI:

Bidding Settings

Set a maximum CPC cap on Maximize Clicks bidding. Claude can set the strategy to Maximize Clicks, but cannot set the “max CPC” limit. You do this in the UI under Bidding Settings. Important: Maximize Clicks is a learning-phase strategy only. Plan to switch to Maximize Conversions or Target CPA after 2-4 weeks once you have conversion data.

Geo Targeting Mode

Toggle between “Presence” and “Presence or interest.” Claude can set which countries to target, but cannot change whether Google shows ads to people physically in those locations versus people who show interest in them. For most B2B campaigns, you want “Presence” only. UI only.

Account-Level Settings

Disable Auto-Apply Recommendations. Google turns this on by default and it will change your campaigns without asking. Claude cannot disable it at account level. You must do this manually in Recommendations -> Auto-Apply.

Ad Creative Limitations

Pin specific headlines to fixed positions in Responsive Search Ads. Claude can create RSAs with multiple headlines and descriptions, but cannot pin Headline 1 to always show first. UI only.

Conversion Tracking

Install or verify conversion tracking tags on your website. Claude can see conversion data in Google Ads, but cannot add the Google Tag to your site or verify it is firing correctly. You must test this yourself with a real form submission.

Campaign Types

Manage Performance Max, Demand Gen, Shopping, or Video campaign creatives at the same depth as Search. The MCP integration is built for Search campaigns. PMax and Demand Gen have different structures that are not fully supported.

URL and Tracking

Set Final URL Suffix or tracking templates. If you use UTM parameters or click tracking, you must add these in the UI. Claude cannot reach them.

Bid Adjustments

Apply device, audience, or schedule bid adjustments. Claude sets the base bid. You manually adjust for mobile, tablet, or specific audiences in the UI.

Extension Types

Add call, lead-form, price, app, or promotion extensions. Claude handles sitelinks, callouts, and structured snippets. The rest are UI only.

Billing

Edit billing, payment methods, or account-level settings. These are account admin features, not campaign management.

The Correct Approach if You Have No Google Ads Experience

This is the exact sequence we recommend. Skip any step at your own risk.

Phase 1: Preparation (15 minutes)

Connect your Google Ads account in Supermetrics token management.
Enable write access in Supermetrics write settings. Do not skip this.
Create a folder on your computer for Claude to save progress files.

Phase 2: Strategy Input (10 minutes)

Give Claude these five things. The quality of your output depends entirely on the quality of this input:

Product/service description: What you sell, who it helps, and what problem it solves.
Buyer persona: Job title, company size, pain points, decision criteria.
Geography: Countries or regions you want to target.
Budget: Daily budget in your local currency.
Conversion event: What counts as a conversion? Form fill? Purchase? Sign-up?

Phase 3: Keyword Research (20 minutes)

Ask Claude to do keyword research and propose an ad group structure. Claude will:

Generate keyword ideas based on your landing page
Pull search volume data for each keyword
Group keywords into themes
Propose an ad group structure

Push back hard here. This is where campaigns fail. If Claude suggests a keyword with 10 monthly searches, reject it. Ask for volume data before approving any keyword. Ask Claude to show you the search volume and competition level for every keyword it proposes.

Phase 4: Landing Page Confirmation (5 minutes)

Before Claude builds anything, confirm:

Which landing pages map to which ad groups
What the conversion event is (and that the tracking is installed)
Whether the landing page loads fast and matches the ad promise

Claude cannot see your landing page quality. You must verify this yourself.

Phase 5: Build (30 minutes)

Let Claude execute the build in one go. Claude will:

Create the campaign
Create ad groups
Add keywords
Write ads
Set bidding strategy
Configure targeting

Claude will return a campaign URL. Open it in Google Ads UI to verify everything looks correct.

Phase 6: Manual UI Fixes (10 minutes)

Apply the three settings Claude cannot reach:

Max CPC cap: Set your maximum cost per click in Bidding Settings.
Geo targeting mode: Switch from “Presence or interest” to “Presence” if you want to target only people physically in your target countries.
Auto-Apply Recommendations: Go to Recommendations -> Auto-Apply and turn everything off.

Phase 7: Conversion Verification (10 minutes)

Test the conversion tag yourself:

Submit a real form or complete a real purchase on your site.
Check Tools -> Conversions in Google Ads.
Verify the conversion shows as “Recorded” within 24 hours.

Phase 8: Wait and Optimize (ongoing)

Wait 2-4 weeks for meaningful data. Then ask Claude to:

Pull the search terms report (what people actually searched for)
Identify negative keywords (irrelevant searches wasting budget)
Propose keyword promotions (high-performing search terms to add as exact match keywords)
Suggest ad copy improvements based on CTR data

What Worked Exceptionally Well

Speed and accuracy. We built a full search campaign with 4 ad groups, 47 keywords, and 12 ads in under 1 hour. The keyword research, grouping logic, and negative keyword sculpting alone would take 2-3 hours manually, and it is easy to make mistakes that drain budget for weeks. The same complete build would take 6-8 hours in the UI for someone who knows Google Ads, or 3-4 days for a beginner.

Audit capability. Claude spotted the Spain geo targeting error in our existing campaign immediately. We would not have noticed this for weeks.

Bulk operations. Adding negative keywords to 4 ad groups simultaneously. Updating ad copy across 12 ads at once. These are tedious in the UI. Trivial for Claude.

Documentation. Claude saved a summary file after each phase. We could stop, resume later, and Claude would pick up exactly where we left off. No lost context.

What Broke (And How We Fixed It)

Write access not enabled. Mid-build, Claude could not create an ad group. Error message was cryptic. Fix: go to Supermetrics write settings and toggle it on. Cost us 10 minutes.

Keyword volume mismatch. Claude suggested keywords with low volume that sounded relevant. We asked for volume data and rejected anything under 100 monthly searches.

Ad copy too generic. Claude’s first draft of ads was technically correct but bland. We edited for our brand voice and added specific numbers (“$49/mo” instead of “affordable pricing”).

The Bottom Line

Claude + Supermetrics MCP is not a replacement for Google Ads knowledge. It is a replacement for the tedious UI navigation that slows down experienced users and overwhelms beginners.

If you understand what a campaign, ad group, and keyword are, this setup lets you execute 10x faster. If you have zero Google Ads knowledge, this setup teaches you as you go, because Claude explains every decision before executing it.

The three UI-only settings are annoying but manageable. The conversion tracking verification is on you regardless of how you build campaigns. The rest, Claude handles.

Our recommendation: Start with a small budget ($20-50/day) and Maximize Clicks bidding to collect data during the learning phase. After 2-4 weeks, switch to Maximize Conversions or Target CPA once you have enough conversion data. Do not try to build a $500/day campaign on day one. The speed of this workflow means you can iterate fast. Use that advantage.

Quick Reference: Claude vs UI

Task	Claude	UI Required
Keyword research	Yes	No
Campaign creation	Yes	No
Ad group creation	Yes	No
Keyword addition	Yes	No
Ad copy writing	Yes (draft)	Final edit
Bid strategy setup	Yes	No
Geo targeting	Yes	No
Negative keywords	Yes	No
Extensions (sitelinks)	Yes	No
Max CPC cap	No	Yes
Geo mode (Presence)	No	Yes
Auto-Apply off	No	Yes
Conversion tag install	No	Yes
Ad headline pinning	No	Yes
PMax campaign setup	Limited	Yes

Ready to Launch Your First Google Ads Campaign?

This guide showed you the exact workflow we used to build a Google Ads search campaign in 1 hour using Claude + Supermetrics. The setup covers 85% of what you need. The remaining 15% (conversion tracking, UI-only settings) is on you.

Next steps:

Connect your Google Ads account in Supermetrics
Enable write access (don’t skip this)
Open Claude and paste your product description, buyer persona, and budget
Let Claude handle the rest
Once your search campaign is live, consider adding AI video ads on social platforms with tools like Arcads or HeyGen for a multi-channel approach

Want more AI advertising guides?

Need the right tools? Browse our reviews and comparisons to find the best AI ad tools for your workflow.

Recommended AI tools for ad campaigns:

Need AI-generated video creatives for your campaigns? Check our Arcads review or SuperScale review
For AI ad copywriting, see our Jasper AI review or the best AI copywriting tools comparison
Running TikTok alongside Google? See our best AI video ad tools roundup

Tools Mentioned

Supermetrics (hub.supermetrics.com) - MCP connector for Google Ads
Claude - AI assistant with tool use capabilities
Google Ads - Advertising platform

Written by Sebastian Wolff. Sebastian is a licensed pharmacist turned digital marketing expert who spends every day testing AI ad tools and teaching companies how to use them.