How Do You Use AI in Google Ads?

Google's own description of Smart Bidding is that it "evaluates billions of combinations of signals to set the right bid for every search query and auction based on your ROI goals". Read the last four words. Your ROI goals are whatever your conversion tracking says they are.

So a double-counted purchase does not make the AI cautious. It makes the AI confident about a number that is twice too big, and it will happily spend to match it. The same is true of a conversion action counting newsletter signups at the same value as an order.

This is why the order in the steps above is not a preference. Handing money to a machine that is reading a broken instrument is the single most expensive thing you can do in a Google Ads account, and it looks like nothing is wrong for about three weeks.

Three things worth being deliberate about.

Auto-apply recommendations, until somebody actually reads the queue. Google is clear that you keep control - "You are always in control, so you can review which recommendations will run on a given day from the queue on the main 'Recommendations' page" - and that "This feature can be turned on or off at any time by updating your auto-apply settings". The gate is real. It just needs a person standing at it.

Generated images of the product itself. Backgrounds, scenes and resizing what you photographed are fine. Inventing a colourway or an accessory is a returns problem and a Merchant Center policy problem at once, and the policy one stops every Shopping ad you have.

Broad match with no negative keyword list. That is not using AI, it is buying searches at random and calling it reach.

Margin. It is last on the list above and it is the one almost nobody does, and the reason is worth saying out loud.

Every other step happens inside the ad account. You can do all of them in an afternoon without leaving Google. This one starts in your own books: what each product cost you, what came back, what is left. That is why it gets postponed forever, and why the list above keeps optimising revenue and calling it performance. Our smart bidding answer explains what that costs you.

Scaley AI does this step for you, which is the only reason it is on this page. It tags every SKU by margin and velocity and syncs the tags into your Merchant Center custom labels, ready to plug into Target ROAS, campaign priority or budget allocation. Nothing is written to the account until you approve it, one item at a time, and every change is logged and reversible. We build Scaley, so weigh that. The track record behind it belongs to ZenoX Media, our sister Google Ads agency, which generated €200M+ in revenue for 200+ ecom brands.

Where you do not need us. If your catalogue is a few dozen SKUs, you can do the margin tagging by hand in a spreadsheet in an afternoon and it will work just as well. If you want your own if-this-then-that rules across many accounts, Optmyzr publishes "Rule Engine-Based Custom Optimizations and Insights" on every plan and Scaley has no rule builder in it. And if you sell services or leads with no product catalogue, the Labelizer has nothing to tag and we are not a fit yet.