Performance Max asks for more image assets than any other Google campaign type, and it punishes missing ones silently — not with a warning, but by declining to enter your ads into auctions on surfaces you did not supply assets for. There is no error message for "you never uploaded a vertical image, so you are not running on YouTube Shorts." You simply do not appear there.
That makes asset coverage a delivery lever rather than a housekeeping task, and it is the reason a PMax campaign with a complete asset group frequently outperforms one with better creative but only the required minimum.
Required, recommended and optional
| Asset | Ratio | Recommended size | Status |
|---|---|---|---|
| Landscape image | 1.91:1 | 1200 × 628 px | Required |
| Square image | 1:1 | 1200 × 1200 px (1080 works) | Required |
| Portrait image | 4:5 | 960 × 1200 px (1080 × 1350 works) | Strongly recommended |
| Vertical / Story | 9:16 | 1080 × 1920 px | Unlocks full-screen surfaces |
| Logo (square) | 1:1 | 1200 × 1200 px | Required |
| Logo (landscape) | 4:1 | 1200 × 300 px | Optional |
Minimums exist — 600 × 314 for landscape, 300 × 300 for square — but assets near the minimum get excluded from large surfaces such as YouTube on connected TVs and Discover. Uploading at the recommended sizes means one file covers every surface rather than qualifying for a subset.
Why coverage beats polish
PMax allocates a single budget across Search, Display, YouTube, Gmail, Maps and Discover. An asset group carrying only the two required images competes for a fraction of that inventory, and the budget concentrates on whatever remains — usually the cheapest, lowest-intent placements.
Google's own asset reporting marks an asset group "Excellent" only with full ratio coverage, and the practical effect advertisers report after adding the 4:5 and 9:16 variants is not better creative performance but access to inventory they were previously excluded from. That distinction matters when you are deciding between another design round and simply completing the set you already have.
Design rules that survive every PMax surface
The same image may render 1200 px wide in Discover and 300 px wide in a Display slot on the same day. That range imposes constraints tighter than any single placement would:
Use one clear subject rather than a composed scene. Keep in-image text minimal, because Google overlays its own headline and description next to your image and competing text reads as clutter. Avoid fine detail that disappears at thumbnail scale. And critically, do not bake in a CTA button — PMax draws its own, and a painted button sitting next to the real one looks like a rendering bug rather than a design choice.
Logos are assets, not decoration
The square logo is required and the 4:1 landscape logo is optional, but the optional one is worth supplying: without it Google pads your square logo into wide slots, which produces visible dead space on either side. Supply both and every surface gets a logo shaped for it.
Reading the asset report without over-reacting
PMax rates individual assets "Low", "Good" or "Best". These labels are comparative within the asset group, not absolute quality judgements — in any group something has to be rated lowest, and replacing it simply promotes a different asset into that slot. Chasing the label produces permanent churn without improving anything.
The rating is worth acting on in two situations. First, when an asset is rated "Low" while visually similar assets are rated "Best", the difference is usually mechanical — wrong ratio for the surfaces it competed on, or text that fails at small sizes — rather than aesthetic. Second, when your whole group sits at "Good" with none at "Best", the group probably lacks the ratio diversity for Google to find a strong performer anywhere.
What the report cannot tell you is which surface an asset underperformed on, which is the single most useful thing you would want to know. Treat asset ratings as a coarse signal and coverage as the lever you actually control.
Images interact with your text assets
Unlike a Meta placement, a PMax ad is assembled at serve time from your images plus your headlines and descriptions, in combinations you do not choose. That has a direct consequence for image design: any message that appears in both the image and a headline will frequently render twice on screen, side by side.
The practical rule is to let the image carry the product and the mood, and let the text assets carry the offer. An image that says "50% OFF" paired with a headline that says "Save 50% today" reads as a templating error to the user, even though both assets were individually correct.
Refresh cadence
PMax gives you no frequency capping you can set directly, which makes asset rotation the only fatigue lever available. Refreshing images every four to six weeks is the common working rhythm. This is another argument for a resizing workflow that is cheap to re-run: if producing a new full set costs a designer two days, the rotation does not happen.
Rotate the images and leave the asset group structure alone. Every structural change — new asset group, changed audience signal — restarts the learning period, so a campaign that is restructured monthly never leaves it. Swapping creative inside a stable group does not.
Producing the full set from one master
The PMax set is exactly the kind of mechanical multi-ratio work that consumes design hours without producing design decisions: the same creative, four ratios, no reinterpretation permitted. Each output has to be recognisably the approved creative, which rules out both centre-cropping and generative reimagining.
AutoAdfit generates the 1:1, 1.91:1, 4:5 and 9:16 set from a single upload at the sizes above, with the subject kept in frame and the original design elements preserved.
Google formats AutoAdfit exports
| Placement | Aspect ratio | Recommended size | AutoAdfit plan |
|---|---|---|---|
| Square | 1:1 | 1080 × 1080 px | Free & Pro |
| Landscape | 1.91:1 | 1200 × 628 px | Free & Pro |
| Portrait | 4:5 | 1080 × 1350 px | Pro |
| Story | 9:16 | 1080 × 1920 px | Pro |