Gemini models are coming to Performance MaxGemini models are coming to Performance MaxGroup Product Manager

Expanded access and improvements to generative AI in Performance Max

AI innovation is delivering powerful results for our advertisers. It’s helping them create more effective campaigns and build creative assets at the volume, velocity and variation required to engage consumers in relevant ways. Marketers and agency partners remain at the heart of the creative process, using a deep understanding of their customers and business to guide AI with the right inputs to build meaningful creative that performs.

In Performance Max, we announced the launch of AI-powered asset generation and image editing in November. These help you generate more text and image assets for your campaign in just a few clicks. Asset generation is now rolling out globally in English with more languages to come. Image editing will finish rolling out in the U.S. in March, with global English expansion following shortly after.

We’re also improving these asset generation capabilities with Gemini, Google’s most capable AI model. Performance Max can now generate long headlines and sitelink generation is coming soon; both of these new features will use the Gemini model’s sophisticated reasoning capabilities to generate text assets.

Soon, when we upgrade our image generation models to Imagen 2, adapted for Google Ads, you’ll be able to generate lifestyle imagery using Performance Max that shows people in action. Similarly, image editing will also include the ability to generate and add backgrounds that feature people. And if you have existing images that are performing well? You’ll be able to generate new options similar to them to scale your creative ideas even further. Of course, advertisers remain in full control and will be able to review all images suggested for their campaigns. And we don’t allow the generation of images of named people, including celebrities and public figures.

In addition, advertisers cannot use Performance Max or asset generation to create content that promotes violence, misrepresentation, or any other content prohibited by our long-standing Google Ads policies and generative AI usage guidelines. We also use mechanisms such as watermarking with SynthID that allow images to be identified as AI-generated. Imagen2 and SynthID are technologies developed by Google DeepMind.

We will continue to improve these new generation capabilities for advertisers in accordance with our AI Principles and practices to continue innovating responsibly.