Home Online Advertising Performance Max Introduces Auto-Generated Text And Image Creative

Performance Max Introduces Auto-Generated Text And Image Creative

SHARE:
Comic: Monkey See, Monkey Generate

Performance Max (PMax), the Google machine learning-based ad product, is notorious for its black box qualities. Advertisers set their preferred KPIs and upload other business priorities and creative assets, then a Google custom bidding algorithm takes it from there.

Soon, you may not even need to upload creative assets. On Tuesday, the company announced the public launch of its auto-generated creative content.

These inputs “steer the AI to goals the [advertisers] care about,” said Brendon Kraham, Google VP of Global Search Ads and Commerce, Business and Product Strategy, during a press conference on Monday.

“Let me take that for you”

PMax’s auto-generated content only does text and images right now, though the images can be added to a video solution that automatically creates campaign assets, said Pallavi Naresh, PMax group product manager.

Text and image creative also get PMax far enough when it comes to advertising that matches the inventory across the Google fleet: Maps, Gmail, Search, Shopping and the online ad network.

One of the common hurdles to scaling campaigns with PMax has been creative bottlenecks. Sometimes, a campaign might call for a specific prompt: an indoor Thanksgiving scene with an autumnal outdoor view, or something of the moment like a person wearing a Travis Kelce jersey.

“Grading, testing and scaling assets can be one of the hardest parts,” said Amanda Silvernale, Google’s global product solutions lead, creatives.

The normal routine of producing an array of images and media, consolidating to a few options, testing them and scaling the winners can be prohibitive for resource-strapped teams. Using a built-in prompt system familiar to any commercial auto-generating chatbot is easy. Plus, there’s a prompt-based editor tool for touching up images.

The obvious pitfalls

Google has “guardrails” in place that prevent the obvious dilemmas that could arise.

The guardrails boil down to adhering to Google Ads policies, which ban using certain keywords, topics or sentiments in ads for protected categories like gambling, health care and alcohol.

Advertisers must also affirmatively select auto-generated creative to be used in campaigns. And

Silvernale said the Google AI product “will never create two identical images, even for the same prompts.”

Any creative generated by the product will also have metadata attached connecting it to the PMax tech, Naresh said, so it can be identified if it’s used elsewhere or by another brand.

The creative is owned by the brand, too, like assets produced by an agency. It can be used elsewhere on or off Google.

Hands-on means hands off

The PMax auto-generated creative is basic for now, but so are the needs of a performance advertising system – if the data is great.

Google’s new PMax search product, Search Themes, launched a week ago and also simplifies the campaign management process by removing advertiser controls and replacing them with general targeting parameters. Just let the Google AI do its thing.

For instance, a museum with fun outdoor entertainment for kids might assign Search Themes like “outdoor recreation” and “kids’ activities,” and then tick the auto-generated content box, rather than staging photos and creating a site landing page about children’s activities.

The product is free, too, Navesh said.

Despite all the positives in its favor, Google treads lightly when it comes to PMax. The company needs agencies and advertisers to trust that the product works when they don’t really know why it chooses any media, audience or format.

“Feedback is and has been critical,” Kraham said in his opening remarks during the press conference.

It’s a more-than-fitting description. While some PMax ad buyers are incensed about the lack of controls, others welcome our machine learning overlords. And, regardless, every single one has changes and information they want to see from the black box.

But as we’ve seen with search keywords, campaign optimization and, now, the first instance of end-to-end machine-made creative on Google Ads, the PMax black box is likelier to continue consuming more advertiser controls then it is to spit something back out.

Must Read

Meta Is Launching An Easy Button For CAPI

Meta is simplifying its CAPI setup and teaching its pixel new tricks, including adding an AI-powered feature that automatically pulls in data from an advertiser’s website.

TelevisaUnivision Joins The Streaming Self-Service Bandwagon

TelevisaUnivision is the latest TV publisher to join the self-serve trend that’s rising in popularity across connected TV advertising. Its streaming inventory is now available to buy through fullthrottle.ai’s self-serve platform. The collaboration includes an ad bidder designed to improve both targeting and measurement.

Comic: Gamechanger (Google lost the DOJ's search antitrust case)

For Google Advertisers Who Overpaid The Monopoly – Don’t Hate, Arbitrate

Law firm Keller Postman is leading mass arbitration suits against Google, seeking advertiser damages for alleged monopoly overpricing. The total available pot is a quarter-trillion dollars.

Privacy! Commerce! Connected TV! Read all about it. Subscribe to AdExchanger Newsletters

Can An AI Solution Fix Misaligned Marketing Orgs?

Opal launched Gem, a new AI solution, to help large brands unify the layers of media and tech within their organizations.

Sports Publisher On3 Tries AI Recommendations To Keep Engagement In Its Home Court

Mula’s AI native content feed helps On3 keep its engagement and RPS consistent amid traffic drop-offs to publisher sites and the growing scarcity of online attention.

Comic: Race To The Bottom

Hearst Built A Unified Ad Marketplace To Simplify Omnichannel News Buys

Hearst is stitching together its far‑flung news properties into a single programmatic marketplace to simplify buying local news and shore up its business as the ad market shifts.