Gemini in Google Ads: Campaigns Start via Chat

Google is pushing generative AI deeper into paid media execution with a Gemini-powered conversational campaign builder in Google Ads. Instead of the classic “fill out 47 fields and pray” setup flow, eligible advertisers now see a prompt-first experience where Gemini drafts campaign structure and assets from a URL and a handful of goals. (Google positions this as a conversational experience for building Search campaigns and related assets; availability can vary by account, language, and policy eligibility.)

This isn’t Google “adding AI.” This is Google quietly redesigning the mental model of ad ops: marketers become editors and approvers, while the platform does more of the assembly work. That’s huge for velocity and a little terrifying if your org still treats governance like a vibe.

What actually shipped

The updated flow reframes campaign creation into a chat-like workflow. You give Google Ads a starting point, typically a business URL (and, depending on campaign type and eligibility, additional inputs like product information), and the conversational experience generates a draft campaign package for you to review.

In practical terms, the system can propose:

• Keyword themes and targeting suggestions aligned to what it infers from your site

• Headlines and descriptions in multiple variants (testing-ready from day one)

• Creative suggestions including image suggestions and, where eligible, generative image variations for certain campaign types

• Negative keyword ideas surfaced as suggestions (negative keyword management itself remains something you review and apply)

• Extensions / assets recommendations (sitelinks, etc., depending on what’s available in your account)

Translation: Google is trying to collapse “research + copy + setup” into “prompt + review.” If you’ve ever launched campaigns during a product drop week, you know why this is appealing.

Why this changes the job

Historically, Google Ads setup punished everyone equally. Small businesses got overwhelmed. Agencies built internal checklists and SOPs just to survive. In-house teams lost days to repetitive setup work (and then still had to chase approvals).

The conversational experience shifts the labor from manual assembly to creative direction + constraints:

• Less time in forms , more time evaluating angles and offers

• More variants by default , which changes how quickly you can test messaging

• Faster iteration loops , because “new draft” is now cheap

This is also Google continuing the broader Performance Max worldview: fewer knobs for humans, more platform intelligence, and a stronger expectation that you’ll feed the machine good inputs (site, product data, creative).

Automation lens: what’s automatable today

Here’s the split: Gemini’s conversational builder is mostly a UI feature. That matters because UI features are great for humans and annoying for automation. If your team’s north star is “plug this into our stack,” the reality is more mixed.

So yes: you can automate Google Ads heavily. But no: you can’t currently treat Gemini’s “first draft campaign” as a callable service the way you can treat, say, an image generation API. That’s the difference between AI assistance and automation infrastructure.

What would need to happen for real API automation

If Google wanted this to become truly stack-friendly, they’d need to expose something like:

• A “draft campaign generator” endpoint (URL or feed plus goals to structured draft output)

• Webhooks or job status (draft ready, draft failed, policy flags triggered)

• Editable “brand + compliance constraints” objects you can pass in (for regulated categories, claims rules, tone, forbidden phrases)

Right now , the platform is acting like the agent, and you’re the human approving actions inside the UI. Later , if Google exposes the agent as an API, your workflow tools (n8n/Make/Zapier/custom) can orchestrate it.

Current vs. future: where teams win (and where they hit walls)

What marketers can do right now

• Launch faster when the UI offers the Gemini flow, especially for Search campaigns where setup time is a real bottleneck

• Increase creative testing volume without needing a copy sprint for every campaign iteration

• Use humans for taste + risk : review drafts, sanity-check intent, remove brand landmines, and align offers to actual business constraints

What still isn’t “automation-ready”

• Headless campaign generation (no official “prompt this campaign into existence” API workflow for the conversational builder itself)

• Governance at scale (you still need your own process for claims verification, approvals, and brand compliance, especially if outputs feel confident but drift)

• Closed-loop creative ops that spans ads to creative storage to reuse to republishing, unless you’ve already built the connectors

Multi-format impact: text, images, video, audio

This isn’t just a text story, even if it starts there.

Text

Gemini is clearly optimizing for rapid copy variant creation. The immediate operational change: copy becomes cheap, so strategy becomes the bottleneck. If your team can’t decide what to test, more variants won’t save you.

Images

Google has been pairing ad creation with generative image capabilities (notably in Performance Max and other AI-assisted creative flows), and it’s reinforcing a new norm: visual variants are becoming part of campaign setup, not a separate design project. That makes small teams more competitive but also increases the risk of low-quality creative entering paid spend unless you have guardrails. (Generative image features have eligibility and policy constraints, and availability varies by region and account.)

Video

Even if the campaign builder is Search-centric in many rollouts, the trajectory points toward multi-format generation across Google inventory. Google has already signaled deeper Gemini integration into Performance Max, including additional text and creative generation features (Gemini models are coming to Performance Max). The direction is clear: the platform wants to help manufacture the asset mix PMax demands.

Audio

Audio isn’t the headline here, and as of now Google Ads does not appear to offer a dedicated, broadly available “generate an audio ad from a prompt” feature inside the Google Ads UI. But audio still fits the same operating shift: once platforms can generate and remix creative quickly (especially for video), audio becomes another variant lane. The missing piece is still the same: publishable, brand-safe outputs with workflow hooks and measurement.

What this means for brands, agencies, and creators

For in-house teams: This reduces time-to-launch, but increases the need for a tight review layer. If you don’t have a lightweight approval system, Gemini just helps you generate more things to argue about.

For agencies: Expect clients to ask why setup takes as long as it used to. Your differentiation shifts to strategy, measurement design, and creative systems, not “we know which checkbox to click.”

For creators and small businesses: This is one of the few AI ad features that can genuinely lower the skill barrier, if your site is clear and your offer is coherent. If your homepage is a confusing brand poem, Gemini will faithfully remix that confusion into ad copy.

The practical reality check

Gemini-assisted campaign building is a real workflow improvement, but it’s not “set it and forget it.” It’s a faster first draft.

The teams that win will treat this as part of a broader human + machine system:

• Machines draft and propose

• Humans decide, constrain, and approve

• APIs execute what’s already structured (today via Google Ads API, tomorrow maybe via Gemini-native endpoints)

Google is turning Google Ads into something closer to a campaign copilot. The automation ceiling is high, but until the conversational builder itself becomes API-addressable, the most scalable move is still the boring one: build a clean ads operating system around the existing API, and use Gemini in the UI where it genuinely saves time.

If you want a bigger picture on how Google is trying to turn websites into campaign inputs across products, see Google Pomelli and Advisors Reshape Campaign Automation.