Create Google Ads and Meta Campaigns in Plain English Using Claude or ChatGPT

Google Ads Manager has 23 distinct steps to launch a Search campaign. Here’s how to do it in one sentence.

That’s not marketing copy — it’s what happens when you connect Claude or ChatGPT to your ad accounts through Adspirer. You describe the campaign you want. The AI builds it. You review it. You approve it. It goes live.

No form-filling. No guessing which setting lives in which submenu. No switching between tabs to cross-reference documentation.

The capability that makes this possible is MCP (Model Context Protocol) — a standard that lets AI assistants call external tools and APIs directly from the chat interface. When you connect Adspirer to Claude or ChatGPT, the AI gains access to 200+ advertising tools that call the Google Ads and Meta Ads APIs on your behalf. You speak. The API responds. The campaign exists.

What “Natural Language Campaign Setup” Actually Means

This isn’t AI generating a campaign plan you then copy into Ads Manager manually. And it isn’t a chatbot that fills in a template form for you.

When you describe a campaign to Claude or ChatGPT through Adspirer, the AI is making direct, authenticated API calls to Google Ads or Meta Ads on your behalf. It reads your description, extracts the structured parameters it needs (budget, locations, keywords, bid strategy, match types, objective), and calls the same APIs that Ads Manager uses in the background.

The difference from working in the UI is that you skip the interface entirely. You write one sentence — or a paragraph with all the details — and the AI handles the translation from human language to platform configuration.

One important clarification: the AI is not “hallucinating” a campaign into existence. It creates real campaigns in your real ad account, using your real budget credentials. The campaigns appear in your Ads Manager immediately. This is why Adspirer defaults every created campaign to paused — so you can review exactly what was built before it starts spending.

You can verify the campaign structure, spot any misunderstandings, and either approve it with one message or ask Claude to change something before it goes live.

What You Need Before Starting

You do not need developer access, API credentials, or any technical setup. Adspirer handles the API authentication — you only authorize via standard OAuth, the same flow you use to connect any third-party app to Google or Facebook.

Step by Step: Google Ads Campaign from One Prompt

Here’s exactly what happens when you describe a Google Search campaign to Claude through Adspirer.

What Claude Actually Builds

When you send that campaign description, Claude creates a complete, layered campaign structure:

• Campaign level — name, objective, bidding strategy, daily budget, status (paused)
• Targeting — geo targets (country or region level), language, ad schedule
• Ad group structure — themed groups matching your keyword intent
• Keywords — each with the match type you specified (exact, phrase, broad)
• API validation — Adspirer checks for policy issues and configuration errors before confirming

What it does not generate automatically: ad copy. Claude can write responsive search ads if you include that in the prompt, but campaign structure and ad copy are separate steps. Most experienced advertisers prefer to handle copy separately anyway.

A Real Confirmation Summary

Here’s what Claude’s pre-build output looks like for the prompt above. This is what you review before saying “go ahead”:

Campaign: "Spring Sale — Running Shoes"
Type: Search
Status: Paused (will not spend until you approve)
Budget: $75/day
Bidding: Maximize Conversions
Locations: United States, Canada
Language: English
Ad schedule: Monday–Friday, 6am–10pm (account timezone)

Ad Groups:
  "Brand Keywords"
    [exact] buy running shoes
    [exact] running shoes sale
  "Generic Keywords"
    [phrase] running shoes
    [phrase] buy running shoes online
    [phrase] best running shoes 2026

Ready to build. Confirm to proceed or let me know what to change.

If the confirmation matches your intent, reply “go ahead” or “build it.” If anything is off — a location is wrong, a match type needs changing, the ad groups need restructuring — correct it in plain English before Claude touches the API. See the full list of skills Adspirer supports →

Google Campaign Types You Can Build This Way

The natural language approach works across all major Google Ads campaign types:

For PMax specifically: describe the objective, geo targets, budget, and audience signals — Claude builds the campaign shell and can pull assets from your landing page URL. Full PMax guide →

Launching a Meta Campaign from Natural Language

Meta campaign creation follows the same pattern: describe the objective, audience, budget, and creative approach, and Claude builds the campaign hierarchy — campaign, ad set, and ad shell — paused for your review.

Other Meta Campaign Types You Can Describe

Beyond image and video ads, the same natural language approach works for:

• Carousel campaigns — describe each card’s headline, link, and description in the prompt
• Collection ads — reference your product catalog by name and Claude sets the catalog source
• Traffic campaigns — specify the URL, audience, placement, and budget
• Awareness campaigns — describe your reach objective, frequency cap, and geo target
• Retargeting campaigns — reference a Custom Audience you’ve already created in Meta (“use my ‘30-day website visitors’ audience”)

The pattern is the same across all types: describe the objective, audience, budget, and any structural specifics — Claude builds the campaign, ad set, and ad shell, paused for review.

What You Can and Can’t Describe

Natural language campaign setup handles most of what you’d do in Ads Manager — but there are limits. Here’s an honest breakdown.

The practical rule: if you can describe it in one or two sentences, Claude can probably build it. If it requires uploading a file or referencing a list you haven’t pre-configured in the platform, handle that part in Ads Manager first, then use Claude for the rest.

According to Google’s official campaign setup documentation, the typical Search campaign workflow involves at minimum six configuration screens before the first keyword is entered. Natural language campaign setup collapses all of this into the prompt itself.

What Happens to Campaigns You Build This Way

Campaigns created through Claude are identical to campaigns built manually in Ads Manager — they live in the same account, appear in the same interface, and can be edited either way. There’s no secondary layer or shadow account. You can build through Claude and then refine in the UI, or vice versa. The only trace of how a campaign was created is in your Claude conversation history.

This also means existing campaigns are editable from Claude. You don’t have to create a campaign through natural language to manage it this way — you can connect Claude to an account with years of existing campaigns and start managing them conversationally from day one.

5 Tips for Better Campaign Descriptions

The quality of what Claude builds is directly proportional to the specificity of what you describe. Here’s how to get it right.

1. Always specify match types explicitly. Don’t say “target these keywords” — say “target these keywords as exact match” or “phrase match.” Without explicit instruction, Claude uses the platform default (broad match for Google, which is rarely what you want). Write out the match type for every keyword or group of keywords.

2. State the status in the prompt. Every prompt that creates a campaign should end with “set to paused for review.” This ensures nothing goes live without your explicit approval. It takes two seconds to type and saves you from accidental spend during setup.

3. Include one description, one campaign. Mixing multiple campaigns in one prompt works but creates risk of misattributed settings. If you’re setting up a Google Search campaign and a Display campaign separately, describe them in separate messages. You’ll get cleaner builds and clearer confirmation summaries.

4. Ask for a pre-build summary on campaigns over $100/day. For higher-budget campaigns, the confirmation step is worth the extra thirty seconds. Send the “list everything you’re about to create” prompt before you say “go ahead.” Review the structured summary carefully — budget, match types, and location targeting are where most errors happen.

5. Use the edit-in-conversation workflow before approving. Don’t approve a campaign and then go into Ads Manager to fix details. If you see something wrong in Claude’s confirmation summary, describe the change in the same chat: “change the bid strategy to Target CPA at $45” or “remove Canada from the locations.” Claude updates via API and confirms the change. You stay in one place.

Bonus: Use a campaign brief template. If you regularly create similar campaigns, write a template description you can copy and fill in each time. Something like: “Create a [campaign type] campaign called ‘[name]’, targeting [keywords/audience], locations [X], budget $[Y]/day, bid strategy [Z], set paused.” Fill the brackets, paste it in, and you have a consistent, repeatable setup workflow that takes 20 seconds.

Comparison: Natural Language vs Manual vs Templates

Who Benefits Most from This Approach

Freelancers and consultants managing multiple client accounts benefit from speed. Building 10 campaigns a month manually takes 3–7 hours of click-work. Describing 10 campaigns takes under 30 minutes, leaving more time for strategy and reporting.

In-house teams unfamiliar with a new platform benefit from the translation layer. If you’ve only run Google Ads and need to launch Meta campaigns, you don’t need to learn a new UI — describe the campaign in terms you understand and Claude handles the Meta-specific configuration.

Agencies onboarding new clients benefit from the audit trail. Every campaign Claude builds is documented in the conversation: what was requested, what was built, what was changed, and when it was approved. That’s a cleaner paper trail than most manual workflows.

Advertisers testing new campaign types benefit from lowered friction. Running a Display campaign for the first time is intimidating in Ads Manager — there are too many placement, audience, and creative options to navigate confidently. Describing what you want and letting Claude handle the configuration removes the interface as a barrier to experimentation.

FAQ

Stop Clicking Through Forms

The ad platform UIs were built for a world where software was the only intermediary between you and the API. That’s no longer true.

When Claude or ChatGPT can call the Google Ads and Meta Ads APIs directly, the UI becomes optional. You still have access to it — for creative uploads, granular bid adjustments, and configurations that require file uploads. But for the majority of campaign setup, one clear description is faster, lower-error, and easier to audit than twenty-three clicks.

The campaigns are real. The data is live. The spend requires your approval. The only thing missing is the form.

For agencies and in-house teams managing multiple accounts: the time savings compound. Building one campaign from a description rather than form-filling saves roughly 20 minutes. Over 10 campaigns a month across 5 clients, that’s more than 16 hours returned to strategy, creative work, and analysis.

The setup takes 2 minutes. The first campaign build takes 60 seconds. Everything after that is just describing what you want.

Natural language campaign setup isn’t a workaround or an approximation of real ad management — it’s direct API access to the same systems Ads Manager uses, with a conversational interface in front of it. The campaigns it builds are indistinguishable from campaigns built manually, because they’re created through the same underlying APIs.

Related Articles

• How to Connect Claude to Google Ads (2026 Guide)
• How to Launch a Google PMax Campaign in 30 Seconds
• PPC Automation with ChatGPT and Claude: The Complete Guide
• How to Run Facebook and Instagram Ads with Claude AI