PMax Asset Groups: 3 Myths Busted for 2026

There’s a shocking amount of bad advice out there about running PMax asset groups, especially with how the AI discovery part works. I see too many marketers still using old-school tactics, which just hobbles their campaigns on a platform that changes practically every week.

Myth 1: You need to create a separate asset group for every keyword or product.

This idea is a holdover from how we used to manage traditional search campaigns, and it’s just wrong for PMax. The AI here is looking at a much bigger picture, trying to find users with high purchase intent across all of Google’s channels instead of just matching a keyword to a headline. When you create tons of tiny asset groups for every single keyword or product, you’re actually fragmenting your data so much that the AI can’t learn efficiently. It needs a consolidated data set to work properly. Instead of breaking things down into tiny pieces, think about grouping products or services that have similar audiences and profit margins. For example, if you sell athletic shoes, you’d be better off with one asset group for “running shoes” and another for “cross-training shoes,” rather than splitting them all up into “men’s running shoes,” “women’s running shoes,” etc. This lets the AI find bigger patterns in user behavior and spend your budget smarter. A 2024 eMarketer report even found that consolidated campaign structures like this were 15% more efficient in conversions than the old, fragmented approach. The machine learning needs enough data volume in each group to work its magic. Too many small groups just dilute the whole process.

Myth 2: The AI will automatically fix poor assets. Quantity over quality is fine.

This belief is flat-out wrong and will absolutely tank your campaign performance. PMax is great at finding an audience, but it’s not a magician. It can’t spin gold from straw. The system is constantly building ads using the headlines, descriptions, images, and videos you give it, so if the ingredients are bland, the final dish will be too, no matter who you serve it to. I’ve seen campaigns where the top 10% of assets drove over 70% of all conversions, while the bottom half did almost nothing but burn impressions and budget. Creative quality is still the biggest single factor in ad effectiveness, accounting for up to 47% of sales lift according to a 2025 Nielsen study, and this applies just as much to AI-driven campaigns. You have to focus on giving the system high-quality and diverse assets. That means you should A/B testing different value propositions and visual styles. Don’t just upload five nearly identical stock photos. Give it lifestyle shots, videos showing the product in use, headlines that focus on benefits, and descriptions that tell a story. The AI needs a rich palette to work with.

Myth 3: Once an asset group is launched, you set it and forget it.

Thinking you can “set it and forget it” with PMax is the fastest way to waste your budget. The entire digital ad space is constantly in flux with changing consumer tastes, market trends, and what your competitors are doing. Your asset groups demand continuous attention. The AI is always learning, but it’s learning from the data you feed it and the market it sees today. If your assets get stale or a new trend pops up that your creative doesn’t speak to, your performance is going to fall off a cliff. Reviewing your asset performance has to be part of your weekly routine. Inside the platform, you can see exactly which headlines, images, and videos are working and which are duds, so look for anything with a “Low” or “Poor” rating and replace it. A good rule of thumb is to refresh about 10-20% of your underperforming assets every month with new creative to test. This keeps things fresh for your audience and gives the AI new things to learn from. And what about seasonality? A holiday campaign needs holiday-specific assets, not the same evergreen creative you’ve been running since July. The IAB’s 2025 “State of Digital Advertising” report confirmed that quick creative iteration is essential for AI campaigns, since their learning cycles are so fast. Because of this, AI reshapes roles on marketing teams, making continuous optimization everyone’s job.

Myth 4: Audience signals are just for targeting. They don’t impact asset group performance.

A lot of advertisers treat audience signals and creative assets like two totally separate things which is a complete misunderstanding of how PMax actually works. Your audience signals, like custom segments, customer match lists, and your own first-party data, are powerful hints for the AI. They tell the algorithm who your ideal customers are, which then influences which combinations of your assets it will show to those specific people. For instance, if your signals point to an audience interested in “eco-friendly products,” the AI will dig through your asset library and prioritize creative that talks about sustainability. If the signals suggest an interest in “premium luxury items,” it will favor assets that look exclusive and high-end. This means that having accurate and updated audience signals directly affects how your creative performs. It’s a feedback loop: strong signals get your ads in front of the right people, and strong assets then get those people to convert.

Myth 5: A/B testing individual assets is the same as A/B testing asset groups.

This is a spot where a lot of smart marketers get tripped up. The PMax AI is already running thousands of micro-tests on its own, constantly combining your assets in different ways to see what works. You aren’t running a traditional A/B test of “ad group A” vs. “ad group B” because the platform is building the ads dynamically. Trying to set up two identical asset groups with one small change to “test” something is a waste of time and will likely just confuse the AI, splitting your budget and learning inefficiently. The right way to do it’s to optimize the individual assets *within a single asset group*. Use the asset reporting to find your underperforming headlines or images. Swap those specific “Low” performing assets out for new challengers. The AI will then fold the new assets into its ongoing testing automatically. If you want a more controlled test, you can use the campaign experiments feature (if it’s available for your campaign type), but the day-to-day work is about improving your creative building blocks. Your job is to feed the system better ingredients, not try to wrestle the steering wheel away from it with old-school testing methods that just don’t apply here. This is even more true as new tools like AI Agents change how we think about UX and tracking.