Last updated: February 14, 2026
Performance Max is the most powerful and most misunderstood campaign type in Google Ads. It runs across Search, Shopping, YouTube, Display, Discover, Gmail, and Maps simultaneously, all from a single campaign. Over a million advertisers use it. Google's own data shows that quality improvements in 2024 alone increased PMax conversions by more than 10% automatically. And yet, the number one question advertisers ask after setting up their first PMax campaign is not about bidding or targeting. It is about creative.
What images should I use? How many headlines do I need? Should I upload my own videos or let Google generate them? Should I split my assets into multiple groups or keep everything in one? How do I know when my creative is getting stale?
These are not trivial questions. In Performance Max, your creative assets are the raw material that Google's AI uses to build every single ad across every channel. The quality, variety, and organization of those assets directly determine how well the algorithm can optimize. Feed it good material and it produces results. Feed it thin, repetitive, or poorly organized material and it produces mediocre ads that waste your budget.
This is the complete guide to Performance Max creative strategy in 2026. We are going to cover exactly what assets you need, how to organize them, how the algorithm selects and combines them, how to read the limited performance data Google gives you, and when and how to refresh your creative. We will also address something most guides ignore: the rapidly shrinking role of human creative decisions in PMax as AI handles more of the assembly process.
Performance Max asset groups accept several categories of creative assets. Here is the full breakdown of what you can (and should) provide.
Images are the visual foundation of your PMax ads. You can upload up to 20 images per asset group. Google accepts three aspect ratios: landscape (1.91:1, recommended 1200x628 pixels), square (1:1, recommended 1200x1200 pixels), and portrait (4:5, recommended 960x1200 pixels). At minimum, you need one landscape and one square image. In practice, you should provide significantly more variety than the minimum. Aim for at least 5 to 8 images per asset group, covering different angles, product shots, lifestyle imagery, and use cases. More variation gives the algorithm more creative options to test across different placements and audiences.
Logos are required for brand identification. Upload your logo in both landscape (4:1 ratio, recommended 1200x300 pixels) and square (1:1 ratio, recommended 1200x1200 pixels) formats. Keep them clean and recognizable at small sizes, because they will appear as tiny overlays on some Display and YouTube placements.
Videos are optional but strongly recommended. You can upload up to 5 videos per asset group (increased to 15 in the January 2026 update). Google accepts landscape, square, and vertical formats, with recommended durations of 10 seconds or longer. Here is the critical detail: if you do not upload your own videos, Google will auto-generate them from your other assets (images, logos, text). Internal testing consistently shows that manually created videos outperform auto-generated ones by 25 to 40%. If you can produce even a simple 15-second branded video, the performance difference is significant. That said, a mediocre auto-generated video is still better than no video at all, because Google's algorithm increasingly favors video-enabled asset groups.
Headlines come in two types. Short headlines (up to 30 characters) are used in more compact ad formats. You can add up to 5 short headlines. Long headlines (up to 90 characters) allow for more descriptive messaging and are used in larger ad formats. You can add up to 5 long headlines as well. Provide the full 5 of each. Each headline should communicate a different value proposition, benefit, or call to action. Do not write 5 variations of the same headline. Write 5 genuinely different messages.
Descriptions can be up to 90 characters each, and you can add up to 5. These work alongside your headlines to provide supporting information. Think of descriptions as the "second line" of your ad that elaborates on whatever headline Google pairs it with. Write descriptions that work in combination with any of your headlines, not descriptions that only make sense paired with one specific headline.
Business name appears across various ad formats and placements. Use your actual business name as customers know it.
Final URL is the landing page where people arrive after clicking your ad. If you enable Final URL expansion, Google may also send traffic to other pages on your site that it determines are more relevant to a specific user's intent. We will cover whether to enable this later in the article.
If you are launching your first Performance Max campaign and do not have a library of professional photography, polished videos, and dozens of headline variations, you can still get started. Here is what the algorithm needs at minimum to function, and what you should aim for to give it a real chance at success.
Google's absolute minimums are 1 landscape image, 1 square image, 1 logo, 3 short headlines, 2 long headlines, and 1 description. Meeting these minimums will get your campaign approved but will severely limit the algorithm's ability to optimize. With so few assets, Google has almost nothing to test. Your ads will look repetitive, your Ad Strength score will be low, and the algorithm will default to showing the same combinations repeatedly because it has no alternatives.
The practical minimum for a functional campaign is closer to 5 images across different ratios and content types, at least 1 video (even a simple one), the full 5 short headlines and 5 long headlines, all 5 descriptions, and your logo in both formats. This gives the algorithm enough raw material to test meaningful combinations and start learning which creative approaches resonate with your audience.
The ideal asset group has 10 to 15 images covering product shots, lifestyle imagery, close-ups, and contextual use cases. It includes 2 to 3 videos in different lengths and formats. All 5 short headlines, 5 long headlines, and 5 descriptions are filled with genuinely varied messaging that covers different angles (promotional, feature-focused, benefit-driven, urgency-based, social proof). This is the asset group that gives the algorithm maximum flexibility and the best shot at strong performance.
If you are using groas to manage your Performance Max campaigns, the platform handles much of this creative optimization process for you. groas evaluates your existing assets, identifies gaps in your asset inventory, and provides guidance on what to add or refresh. It continuously monitors asset performance ratings and flags when creative fatigue is setting in, which we will discuss in detail shortly.
This is one of the most debated questions in Performance Max strategy, and it comes up constantly in search data. Should you put all your images and assets into a single asset group and let the algorithm sort it out? Or should you create separate asset groups organized by theme, product category, or audience?
The answer depends on your business and your conversion volume, but here is the framework that works for most advertisers.
Single asset group approach works best when you have a single core product or service, your monthly conversion volume is low to moderate (under 50 conversions per month), and you want to keep things simple while maximizing the data available to the algorithm. With one asset group, all your conversion data feeds into a single optimization pool. The algorithm has the maximum amount of data to work with, and it can test any combination of your assets against any audience. The downside is that you lose thematic relevance. A headline about "free shipping" might get paired with an image that has nothing to do with shipping, or a product-specific description might show alongside a generic lifestyle photo.
Multiple themed asset groups work best when you sell multiple distinct product categories or services, your monthly conversion volume is high enough to support multiple groups (ideally 30+ conversions per month per group), and thematic relevance between images, headlines, and landing pages matters for your conversion rate. For example, an ecommerce store selling both running shoes and hiking boots would benefit from separate asset groups because the imagery, messaging, and landing pages are fundamentally different. A law firm offering both personal injury and family law services would similarly benefit from separate groups because the audience, intent, and creative approach are completely different.
The key constraint is data. Each asset group needs enough conversions to learn independently. If you split your campaign into 5 asset groups and each one only gets 5 conversions per month, none of them will have enough data for the algorithm to optimize effectively. Start with 1 to 2 asset groups and only split further when your conversion volume supports it.
When testing creative approaches (product photography vs. lifestyle imagery, promotional messaging vs. benefit-focused messaging), the best practice since January 2026 is to use Performance Max's built-in A/B testing feature for creative assets. This lets you run controlled split tests within a single asset group, comparing two distinct asset sets without needing to create separate campaigns or duplicate your asset groups. Run tests for at least 4 weeks to account for the learning phase and gather statistically meaningful data.
Performance Max does not show a single predetermined ad. Every impression is a unique combination of your assets, dynamically assembled in real time based on the context of that specific user, placement, and moment.
Here is what happens behind the scenes. A user triggers an ad opportunity (through a search query, YouTube viewing behavior, Display Network browsing, or email check in Gmail). Google's algorithm evaluates the available ad space and format requirements. Based on the placement (YouTube pre-roll requires video, Search requires text, Display requires images), the algorithm selects compatible assets from your group. It then chooses the specific combination of headline, description, image or video, and call-to-action that it predicts will perform best for this particular user based on their device, location, time of day, browsing history, and the signals it has learned from your campaign data.
This means your YouTube ad might use Video A with Headline 3, while your Search ad uses Headline 7 with Description 2, while your Display ad uses Image 5 with Headline 1 and Description 4. Every combination is tested, and the algorithm progressively shifts impressions toward the combinations that produce the best results.
This dynamic assembly is why asset variety matters so much. If all your headlines say essentially the same thing with different wording, the algorithm has no meaningful variation to test. If all your images are product shots on white backgrounds, the algorithm cannot discover that lifestyle imagery performs 28% better on YouTube (which is a common finding). The algorithm is only as good as the raw materials you give it.
There are a few patterns worth understanding about how the algorithm tends to prioritize assets. First, it leans heavily toward assets that drive early conversions. An image that generates a few quick conversions in the first week will get more impressions going forward, even if another image might ultimately perform better with more data. This "rich get richer" dynamic means your initial asset selection has an outsized impact on long-term performance.
Second, the algorithm generally gives more impressions to asset groups that have more complete asset inventories. Groups with all asset types filled (images, videos, text) tend to receive more distribution than groups with missing asset types, because they are eligible for more placement formats.
Third, video assets tend to unlock significant additional reach. Google's algorithm favors video-enabled campaigns for YouTube and Discover placements, and these channels often drive incremental conversions that image-only campaigns miss entirely.
One of the most frustrating aspects of Performance Max is the gap between how much the algorithm knows and how much it tells you. Google assigns each asset a performance rating: Best, Good, or Low. It also provides an overall Ad Strength score for each asset group (Poor, Average, Good, Excellent). And as of 2025, the asset report shows impressions, clicks, cost, and conversions for each individual creative element.
These metrics are useful but have significant limitations. The performance rating (Best, Good, Low) is a relative comparison within your asset group, not an absolute measure of quality. An asset rated "Best" in a group of mediocre assets might still underperform an asset rated "Good" in a stronger group. The rating tells you which of your assets Google prefers relative to each other, not whether your assets are objectively good.
The impressions and clicks data at the asset level is helpful for understanding which assets get the most distribution, but it does not tell you which specific combinations of assets drive conversions. You might see that Image 3 has the most impressions and Headline 5 has the highest click-through rate, but you cannot see whether Image 3 paired with Headline 5 specifically is what drives results, or whether a different combination is actually doing the heavy lifting.
Here is how to work with what you have. Check your asset performance ratings weekly, but do not make changes based on fewer than 14 days of data. Any asset rated "Low" after 2 to 3 weeks of sufficient impression volume is a candidate for replacement. Assets rated "Best" should be protected and kept in your group. Assets rated "Good" are worth keeping but should inspire you to create new variations that test different approaches. The Ad Strength score is a directional indicator. Getting from "Average" to "Good" or "Good" to "Excellent" by adding more asset variety is almost always worth the effort. Going from "Excellent" to... still "Excellent" by adding marginal assets of questionable quality is not.
For deeper creative insights, look at your performance data at the asset group level rather than the individual asset level. If one asset group consistently outperforms another, the creative theme of that group is resonating. Use that learning to inform your creative direction for future assets.
Creative fatigue affects Performance Max campaigns just like any other advertising channel, but it manifests differently because the algorithm is constantly recombining your assets. The fatigue is not "people have seen this exact ad too many times." It is "people have seen variations of this same creative approach too many times, and the novelty has worn off."
The telltale signs of creative fatigue in PMax include a gradual decline in click-through rate over 3 to 4 weeks (not a sudden drop, which usually indicates a different problem), increasing cost per conversion without any change to your targeting or budget, a flattening of conversion volume where previously there was growth, and more of your assets being downgraded from "Best" or "Good" to "Low" in the performance ratings.
For most advertisers, the right refresh cadence looks something like this. Review asset performance weekly. This does not mean making changes weekly. It means looking at the data so you can spot trends early. Replace underperforming assets every 2 to 3 weeks. When an asset has been rated "Low" for 14 or more days with meaningful impression volume, swap it for a new variation. Do not replace more than 2 to 3 assets at once to avoid destabilizing the asset group's optimization. Conduct a major creative refresh every 4 to 6 weeks. This is where you introduce new creative themes, seasonal messaging, updated product imagery, or entirely different visual approaches. Plan these around your business cycles (new product launches, seasonal promotions, major campaigns).
One critical warning: do not refresh all your assets simultaneously. If you remove every image and headline and replace them with entirely new ones, you are essentially starting from scratch. The algorithm loses all the performance data associated with your previous assets and has to relearn everything. Treat creative refreshes like ship-of-Theseus updates. Replace a few planks at a time so the ship keeps sailing while gradually becoming something new.
groas automates much of this creative monitoring process. The platform continuously evaluates asset performance across your PMax campaigns, identifies assets that are showing signs of fatigue, and alerts you when it is time to refresh specific elements. Rather than requiring you to manually check performance ratings every week, groas surfaces the creative decisions that actually need your attention and provides data-backed recommendations for what to change.
Here is the honest conversation that most Performance Max guides avoid. The human role in PMax creative is shrinking, and it is going to keep shrinking.
In 2023, a skilled PPC manager would carefully craft each headline, hand-select every image, agonize over ad copy variations, and spend hours thinking about which creative would work best in which placement. In 2026, Google's algorithm makes those combination and placement decisions far better than any human can, because it processes data from millions of impressions across billions of users to determine which creative approach works best for each specific individual in each specific context.
Google's Asset Studio, powered by Imagen and Veo AI models, can now generate images and videos directly within the Google Ads interface. The brand guidelines feature introduced in 2025 ensures AI-generated assets maintain your visual identity. Automatically created text assets scrape your landing pages and domain to generate additional headlines and descriptions. The A/B testing feature runs controlled experiments that would have taken weeks of manual setup.
What does this mean for the advertiser? Your creative role in PMax is shifting from "design the ad" to "supply the building blocks." You are less like an art director and more like a content provider. Your job is to ensure the algorithm has high-quality raw materials to work with: strong product photography, genuine lifestyle imagery, clear and varied messaging, and (ideally) video content that showcases your product or service. Once those materials are in the system, your job is to monitor what the algorithm does with them and refresh the inventory when it gets stale.
This is not a loss of control. It is a reallocation of effort toward the things that actually matter. The creative decisions that still require a human brain are high-level ones: what is our brand message, what differentiates us from competitors, what emotional response do we want to evoke, what new products or promotions should we feature? The tactical decisions (which headline pairs with which image on which placement for which user at which time of day) are now handled better by AI than by any human.
This shift maps perfectly onto the broader move toward autonomous Google Ads management. As creative assembly becomes increasingly automated, the remaining human creative inputs become something that tools like groas can facilitate. groas monitors your asset performance continuously, identifies when the algorithm needs fresh material, and guides you on exactly what kind of creative to provide. The human time investment shrinks from hours per week to minutes per month, focused entirely on the strategic creative decisions that still require human judgment.
If you are launching a new PMax campaign, start with a single asset group containing at least 5 to 8 images across all three aspect ratios, at least 1 custom video (even a simple 15-second branded clip), all 5 short headlines, 5 long headlines, and 5 descriptions with genuinely varied messaging, your logo in both formats, and a strong primary landing page. Get your Ad Strength to "Good" or "Excellent" before launching. Only split into multiple asset groups when your conversion volume supports it (30+ conversions per month per group).
If you are optimizing an existing PMax campaign, run an asset audit. Check which assets are rated "Low" and have been for more than 2 weeks. Replace them with new variations, not more of the same thing. If you do not have videos, create or upload at least one. Ensure you have all asset slots filled to maximum. Consider using the A/B testing feature to test fundamentally different creative approaches (product photography vs. lifestyle, promotional messaging vs. educational).
If your PMax performance is declining, check for creative fatigue before changing anything structural. A campaign that has been running the same assets for 6 or more weeks likely needs a creative refresh more than a targeting change or budget adjustment. Replace 2 to 3 assets at a time. Monitor performance for 2 weeks after each change. If performance does not recover, try introducing entirely new creative themes rather than incremental variations on what you already have.
Google allows up to 20 images per asset group. The minimum is 1 landscape and 1 square image, but aim for at least 5 to 8 images covering product shots, lifestyle imagery, and different use cases. Provide images in all three aspect ratios (landscape 1.91:1, square 1:1, and portrait 4:5) to ensure your ads can appear in every available placement format. More variety gives the algorithm more combinations to test and generally produces better results.
Start with 1 to 2 asset groups and only split further when your conversion volume supports it. Each asset group needs enough data to learn independently, ideally 30 or more conversions per month. Use multiple asset groups when you sell distinct product categories or services that require different imagery, messaging, and landing pages. If your total campaign generates fewer than 50 conversions per month, a single asset group will almost always outperform multiple groups because it concentrates all your data in one optimization pool.
If you have a single product or service, mix them in one group. The algorithm will test which image styles perform best across placements and audiences. If you have distinct product lines with different audiences (running shoes vs. hiking boots, personal injury law vs. family law), use separate themed asset groups so the images, headlines, and landing pages all stay thematically aligned. For testing whether product photography or lifestyle imagery performs better within a single product category, use the Performance Max A/B testing feature introduced broadly in January 2026.
Google will auto-generate videos from your existing images, logos, and text. These auto-generated videos are functional but typically underperform manually created videos by 25 to 40% based on internal testing data. If you can produce even a basic 15-second branded video, do it. A simple product showcase with text overlays and your logo is dramatically better than Google's auto-generated slideshows. Google's algorithm gives more impression distribution to video-enabled asset groups, so not having custom video means missing out on reach as well as performance.
Review asset performance ratings weekly. Replace individual assets rated "Low" every 2 to 3 weeks. Conduct a broader creative refresh (new themes, seasonal messaging, updated product imagery) every 4 to 6 weeks for static assets and every 8 to 12 weeks for video. Never replace all assets simultaneously, as this resets the algorithm's performance data. Replace 2 to 3 assets at a time and allow 2 weeks between changes for the algorithm to stabilize.
groas continuously monitors asset performance across your PMax campaigns, tracking performance ratings, impression distribution, and signs of creative fatigue. Instead of requiring you to manually check asset reports every week, groas identifies which specific assets need attention and when. It flags underperforming assets, recommends refresh timing, and provides guidance on what types of creative to add based on gaps in your current asset inventory. This turns PMax creative management from a recurring manual task into an occasional, focused activity where you provide fresh building blocks when the data indicates they are needed.
Ad Strength measures the variety and completeness of your asset group on a scale from Poor to Excellent. It is a directional indicator, not a performance guarantee. Campaigns with "Excellent" Ad Strength have more asset variety for the algorithm to test, which generally leads to better performance. Aim for at least "Good" and ideally "Excellent" by filling all asset slots with diverse, high-quality content. However, do not add low-quality filler assets just to boost the score. Five excellent images with a "Good" Ad Strength will outperform 20 mediocre images with an "Excellent" score.
Less so than it used to be. Through 2024 and 2025, Google introduced asset-level reporting showing impressions, clicks, cost, and conversions for individual creative elements, channel-level performance reporting, expanded search term visibility, and A/B testing for creative assets. You still cannot see exactly which combinations of assets perform best together, and you cannot manually control which assets show on which channels. But the level of creative visibility is dramatically improved from PMax's early days. The remaining reporting gaps are one reason tools like groas are valuable, because they can continuously analyze the available data and surface creative insights that would take hours of manual analysis.
