Google Ads gives marketers tighter rein over AI campaigns

Running AI ad campaigns has always come with an unspoken anxiety: the automation gets results, but you have no clue why, and the machine-generated visuals might completely butcher your brand guidelines.

On September 18 at the DMEXCO marketing conference, Google introduced a hefty batch of new features for Google Ads. The theme was unmistakable. While Google is doubling down on Gemini models to generate ads, it is finally giving marketers the manual steering wheel they have been begging for.

Guardrails for AI creatives

Performance Max (PMax) has long driven conversions across Google inventory, but marketers have constantly complained about its black-box nature. Google is finally addressing this with strict creative guardrails.

Advertisers can now upload up to five reference images alongside a text prompt. Instead of hallucinating arbitrary styles, the AI uses these references to mimic the exact visual aesthetic of the brand. Starting next month, brand guidelines in PMax will roll out to general availability, letting teams lock down specific fonts, color palettes, and logos.

Generative image editing is also moving beyond PMax into Search, Demand Gen, App, and Display campaigns, including direct editing of product photos from Google Merchant Center.

On the copy side, Google is expanding its conversational campaign builder powered by Gemini into German, French, and Spanish. According to Google, small businesses using this conversational workflow were 63% more likely to publish Search campaigns with Good or Excellent Ad Strength. Full asset generation across major campaign types is also adding six languages, including German, French, Spanish, Portuguese, Dutch, and Italian.

Peeking inside the PMax black box

Better controls are nice, but campaign transparency was the real headline for performance media buyers.

The most requested feature in PMax history, campaign-level negative keywords, is finally rolling out as a beta by the end of the year. Previously, keeping automated campaigns away from unwanted search queries required awkward workarounds or Google support reps.

Reporting is getting a major overhaul as well. Starting next month, PMax will introduce asset-level conversion metrics and asset coverage reports that serve up blunt advice, like telling you to add two long headlines or three square images. It is also introducing Impression Share reporting for the first time in PMax, showing exact visibility across Search and Shopping inventory.

Demand Gen campaigns also received updates. Creative preferences are now available to all advertisers, allowing teams to pin specific videos to designated placements, with omnichannel bidding for online and in-store conversions entering beta soon.

Why it matters

For performance marketers and business owners, these tools are worth testing immediately. Uploading reference images and locking brand guidelines solve the messy aesthetic problem that kept creative directors away from AI assets. Meanwhile, campaign-level negative keywords will directly prevent wasted spend on irrelevant queries.

Across the ad tech landscape, this marks a shift in the automation narrative. For years, platforms insisted on total end-to-end algorithmic control. Google even cited a TransUnion study showing PMax delivered 19% higher ROAS in 2023 than automated campaigns on the largest social platform. But enterprise advertisers simply refused to hand over full control to an unchecked algorithm.

In our view, Google did the right thing by taking a step back from pure black-box automation. Pairing generative scale with strict brand controls and negative keywords offers the exact hybrid setup marketers actually want.