Google Ads AI: what Google's own AI does, and what it leaves to you

Google uses AI in Google Ads for three jobs: setting bids, choosing where ads show and writing parts of the ads. The features below carry Google's current names, checked on Google Ads Help. Google renames them often: automatically created assets are now text customization.

Every one of them optimizes toward the conversion goals and the targets you give it. That makes the inputs the real decisions. A conversion action that counts the wrong event trains all of them on the wrong signal, and a target set too loose is one Google will happily meet.

Google defines Smart Bidding as bid strategies that use Google AI to optimize for conversions or conversion value in every auction, which it calls auction-time bidding. There are four: Target CPA, Target ROAS, Maximize conversions and Maximize conversion value. They read signals a person cannot weigh per auction, such as device, location, time of day, remarketing lists and the actual search query.

Google recommends judging a Smart Bidding strategy over a period with at least 30 conversions, such as a month, or 50 for Target ROAS. After a new strategy or a setting change, the strategy shows Learning, which Google says typically lasts one to two conversion cycles.

Performance Max is a goal-based campaign type that reaches YouTube, Display, Search, Discover, Gmail and Maps from a single campaign. You supply the conversion goals, the budget, an optional CPA or ROAS target, text, images and videos, and audience signals that guide its learning.

Google then decides which channel, which placement and which combination of assets each impression gets. Reporting now includes channel performance, asset group and placement reports, but the split between channels remains Google's decision rather than a setting.

AI Max for Search campaigns is not a campaign type. It is a campaign-level setting on a Search campaign with two parts. Search term matching extends your keywords with broad match and keywordless technology. Asset optimization adds text customization, which writes headlines and descriptions from your landing pages, domain, existing ads and keywords, and Final URL expansion, which can send a search to a more relevant page of your site.

Google's FAQ says AI Max relies on automated bidding such as Maximize conversions or Maximize conversion value, and that search term matching does not work with manual CPC. The controls remain: brand inclusions and exclusions, locations of interest, URL inclusions and exclusions, and text guidelines. In the search terms report, the searches it found show AI Max as their match type.

The Recommendations page suggests changes and scores the account. Auto-apply lets you opt in to bundles such as Maintain your ads and Grow your business, which Google then applies for you; the History tab lists what was applied, and any recommendation can be switched off again.

Ask Advisor, in beta, is Google's agentic conversational experience in Google Ads, built on Gemini. It answers questions, troubleshoots performance and policy issues, and suggests new text and image assets that you review before they apply. Google lists it as unavailable in manager accounts, so it works one account at a time.

Each feature optimizes one Google Ads account toward the target you set. None of them knows whether that target still leaves a profit after product costs, shipping and returns, because Google does not see your margin. None of them compares Google with Meta, or reads what your store recorded rather than what Google attributed.

The decisions with the most money behind them stay with you by design: the conversion goal, the target CPA or ROAS, the budget and the campaign structure. Those are the changes worth a second look before they run.

• Is the target CPA still below what an order contributes?
• Is Google or Meta paying back better, on store revenue?
• Which of several accounts needs attention first?
• Who made the last budget change, and what happened after it?

Adrails works above Google's AI rather than in place of it. It reads Google Ads and Meta accounts side by side, with Shopify or Stripe revenue next to the spend. Its agent, or Claude and ChatGPT through the Adrails MCP server, prepares changes to the inputs Google optimizes toward: a budget move of at most 30%, a pause or a resume, a new campaign created paused, a new target CPA or ROAS.

Each change is a proposal that an owner or admin applies in Adrails. When it is applied, Adrails reads Google Ads again and blocks the change if the campaign moved in the meantime; pauses, resumes and budget changes are then observed after 24 hours, 3 days and 7 days. Smart Bidding still sets every bid.