Google has rolled out their new Performance-Max AI Generation Tools in Beta. In this article I’ll review and rate the new features and let you know whether these are valuable editions to your creative toolbox.

What are the new tools?

Tools for creating headlines and descriptions build upon what Google was already doing with automatically generated ad copies. Instead of just suggesting ad copy from a list of seed types, Google now allows you to interact with the AI by describing your offer and unique value proposition. Additionally, whereas the older automation uses landing page content and other account ads, the AI tool bases copy on its understanding of your products and services. Is that really different? Probably not much but in theory this should yield better results.

The new tool for creating images is similar to other tools of the same ilk. You type in what you want and what you don’t want, and Google whips it up from nothing.

Creating headlines and descriptions

To get started just click on the “Generate headlines” button. You’ll be presented with a new interface window with three tabs along with a list of your currently used headlines or descriptions.

The first tab is the old school “Suggested” content. It’ll show you content from your landing pages and other ads and also shows you recently used creatives. The big problem that has always existed with this is it’ll pull from old irrelevant ads you haven’t run in years. And if you run dynamic ads you will receive all manner of garbage suggestions. We still have accounts that pop up recommended headlines and descriptions that contain phone numbers. Add these and your ad will be instantly disapproved.  The old ads being referenced are from very many years ago when Google did allow you to insert phone numbers directly into your creatives.

The second tab displays your asset library. You can easily scroll through this and select ad copies to add to your ad. But in practice the layout is hard to work with. If you have hundreds of headlines, for example, you’ll be scrolling through pages for ages to find what you’re looking for. This particular tool would benefit from filtering and sorting tools. Google did add a filter for choosing asset library folders, but let’s be honest, no advertiser has ever properly organized their asset library neatly into folders.

The third tab “Generated” is where the magic is supposed to happen. Google will often present some default ideas which could be pretty good or horrendous… mostly they tilt to the latter in my experience so far. Not only is the creative lacklustre to be kind, but it’s often not even aligned with what’s being offered. Now you have the option to enter your UVP or offer in the text box provided. Unfortunately, even when you succinctly spell out how you crush your competitors all Google does is return duplicate content or copy that’s bland or off topic.

performance-max ai

Honestly, I’ve been able to generate way better ad copy variations using simple prompts with ChatGPT or Gemini directly. What’s up Google?

Creating Images

The other big tool in the new arsenal is the automatic image generator. As with the new headline/description generation tool, the screen will present the legacy options suggested, asset library, website/social, and free stock plus the new “generated” tool.

This works similar to DALL-E 2 with just basic text box entry. Google doesn’t provide any granular controls such as the model type or negatives but you can choose a specific aspect ratio. The images Google cranks out are high quality and ad ready.

google image generator

But I’ve found that with a lot of images you only get a useable square version by default. When you try to crop the image for landscape or portrait it’s not workable.

suv with accessories

Google does offer a filter to select the image ratio you want. But then, of course, the other two aspect ratios don’t always work. And since you won’t get the same exact truck (per this example) in separate operations, in this example, you will end up with banners showing a different truck. This is far from ideal assuming you want to generate all 3 aspect ratios for your creative.

Besides this issue, there are two massive restrictions!

First, the tool will not generate any images of people or faces. This is stated directly in the interface and relates to the recent debacle over Gemini generating racially biased images. Second, the tool will not generate any images based on brands. If you try to include a brand Google will reject your prompt. For example, if I replaced “SUV” with “Jeep CJ” in my prompt for the image above Google will not accept it.

In combination this makes the tool practically useless for most advertisers.

And this is really too bad, because it would be great to have a robust built-in image generator that spits out everything in the correct aspect ratios. Of course, humans/faces generation problem may well get fixed shortly. But the “brand” limitation is a problem for any company that wants to generate product images featuring their own brand or their own products in use with 3rd party brands, i.e. cottage industry items like Apple iPhone cases. And not being able to get all 3 aspect ratios from a single image is a big problem too.


The new P-Max AI content generation tools are interesting but not ready for prime time. The headlines and descriptions are often out to lunch, and you can do much better using Gemini directly. The image generator produces photo-quality images but there currently is no way to create humans/faces which is a giant drawback. In addition, you can’t include brands in image generations. That one-two combination of restrictions makes this tool all but useless for most marketers. Worse, you cannot easily generate a single image that usable in all aspect-ratios.