Build and Launch Google Search Campaigns by Chat

You can now build and launch Google Search campaigns by chat — keyword research, competitor intel, campaign creation, and ongoing budget management, all without opening the Google Ads interface once. AdWhispr's MCP server ships Google Search execution as generally available, and it plugs into Claude, ChatGPT, Claude Code, or Cursor.

This is a tutorial. By the end you'll have run the full loop: research the keywords, see what competitors bid on, launch the campaign, and manage it — from one conversation.

Why Search campaigns by chat actually make sense

Search is the most structured ad format there is. A Search campaign is keywords, match types, ad copy, and a budget. There's no creative production bottleneck, no video shoot, no design review. Which means the entire thing can be specified in plain English — and an AI assistant with the right tools can build it correctly.

The catch has always been the "right tools" part. A chat assistant with no keyword data guesses. A chat assistant with execution access but no research just launches your guesses faster. What you want is research and execution in the same session, so the campaign you launch is built on data, not vibes. That's the whole design behind AdWhispr's MCP server.

Step 1: Connect your Google Ads account

First, get AdWhispr into your assistant — the setup for Claude, ChatGPT, Claude Code, and Cursor is documented at adwhispr.com/integrations (we also cover Claude Code and Cursor setup in detail). Then connect your ad account:

"Connect my Google Ads account"

That fires connect_ad_account, which walks you through a standard OAuth flow. Once you're connected, list_ad_accounts shows every account you can launch into. No API tokens to paste, no developer console.

Step 2: Research keywords — yours and your competitors'

Two research tools feed a Search campaign:

research_keywords — keyword intel for your own product space:

"Research keywords for a DTC cold brew coffee subscription.
What should a Search campaign target?"

research_competitor_keywords — what your competitors actually bid on. This is the step most people skip because it used to require a separate tool and a separate subscription:

"What keywords is [competitor] bidding on in Google Search?
Which ones overlap with mine, and which are they on that I'm not?"

The competitor angle matters more than most buyers admit. Your own keyword brainstorm reflects how you describe your product. Your competitors' bids reflect what's actually converting in your category — they're spending real money on those terms, week after week. It's the Search equivalent of our days-running performance proxy: sustained spend is the signal, because nobody keeps paying for keywords that don't work.

Because this is a conversation, you can iterate before anything launches:

"Drop the branded terms, keep the high-intent ones, and group them
into two ad groups: 'cold brew subscription' and 'coffee delivery'."

Step 3: Launch the campaign

When the keyword set looks right, launch it with launch_search_campaign:

"Launch a Google Search campaign with those two ad groups.
$40/day budget, targeting the US. Write the headlines and
descriptions from my brand's positioning."

Your assistant assembles the campaign — keywords, ad groups, responsive search ad copy, budget — and shows you the plan before it goes live. You approve in chat. Done. launch_search_campaign is generally available today, alongside Performance Max on the Google side and TikTok execution.

Step 4: Manage it without leaving the chat

Launching is half the job. The management tools close the rest of the loop:

A realistic Monday-morning check-in looks like this:

"Pull performance for my Google account for the last 7 days.
If anything is clearly underperforming, tell me — and suggest
where the budget should move."

Then act on the answer in the same breath:

"Do it. Pause the weak one and move its budget to the winner."

This is where flat pricing matters. AdWhispr's Pro plan is $39/mo with unlimited research tool calls — not metered per call. Execution-only MCP servers in this category typically charge per tool call, which quietly punishes exactly this kind of iterative back-and-forth. A keyword research session where you refine the list five times shouldn't feel like a taxi meter. (We've written about how we compare to Adspirer, the best-known metered option, if you want the honest breakdown.)

The part legacy tools can't do: research and launch in one loop

Here's the full session, end to end, as one workflow:

1. research_keywords — map your product's keyword space.
2. research_competitor_keywords — see what rivals actually bid on, and steal the overlap analysis.
3. Refine the list conversationally — no spreadsheet exports.
4. launch_search_campaign — launch with budget and geo in one prompt.
5. get_account_performance + update_budget + pause_campaign — manage weekly by chat.

Keyword tools can do step 1. Spy tools can sort of do step 2. Execution servers can do steps 4–5. AdWhispr is the only MCP server we know of where all five happen in one conversation, with the research feeding the launch directly — no copy-paste between tools, no strategy left behind in a browser tab.

And Search is just one lane. The same connected account and the same session can launch Performance Max campaigns, and the same MCP server does TikTok research and execution too.

Try it on your next campaign

The free tier gives you 5 MCP tool calls a month — enough to connect your account, run a keyword research pass, and see the loop work before you commit. Pro is $39/mo with a 3-day free trial and unlimited research calls.

Setup instructions for Claude, ChatGPT, Claude Code, and Cursor are at adwhispr.com/integrations. Your next Search campaign can be a conversation.