What you'll learn in this article
- Which performance max specs are hard limits the system will reject, and which are "recommended" numbers you can miss without an upload ever failing
- The exact minimum set of assets I upload to get an asset group live and out of "limited" ad strength, from accounts I run every day
- Why an asset can pass every dimension check and still be silently dropped from serving
- What I actually see move creative ranking inside PMax once the specs are met, and what looks like it should matter but doesn't
- The logo and video traps that quietly cap your ad strength without ever showing an error
The performance max specs table everyone links to tells you the dimensions and the character counts. What it doesn't tell you is the thing that actually matters in a live account: which of those numbers the system enforces with a hard rejection, which it treats as a polite suggestion, and which asset passes every check on upload and then never serves. I have inherited enough asset groups sitting at "limited" ad strength with a full green checklist to know the spec sheet is the start of the job, not the end of it.
So this is not a reprint of the reference table. Google's own Google Ads help page has every dimension for a performance max campaign, and I link it below so you have the source of truth for the raw numbers. What I want to cover instead is the layer underneath: the performance max ad specs that cause an outright reject, the real floor for escaping "limited," and the handful of creative signals I have watched genuinely change how image assets and video assets rank once everything technically valid is in place. The aspect ratios and file rules are only the surface. All of it comes from accounts I manage, not from a help-center paraphrase.
If you are still deciding whether this campaign type is the right container for your spend before you obsess over asset dimensions, my breakdown of what Performance Max is and how it fits an account is the better starting point; here I assume the campaign exists and the only open question is what to upload and why.
The performance max specs the system actually rejects
The first thing to separate is a hard reject from a soft "recommended." Most of the numbers in the reference table are recommendations for ad strength; only a subset will actually stop an asset from going in. Knowing which is which saves you from chasing spec compliance that changes nothing while ignoring the one that gets your asset thrown out.
Images: dimension floors and the file-size wall
On images, the rejections I actually see cluster around three things. File type first: JPG and PNG go in, anything else bounces at upload. File size second: the ceiling is 5120 KB, and oversized files fail silently often enough that on any account with heavy hero images I compress before I ever open the editor. Third, the minimum dimensions are a genuine floor, not a suggestion a landscape below 600 x 314 or a square below 300 x 300 will not upload. The recommended 1200 x 628 and 1200 x 1200 are a different category: miss those and the asset still serves, it just serves smaller and tends to get out-ranked. I treat the minimums as pass/fail and the recommended sizes as the real target, because that is how the system treats them too. My note on how PMax rates and scores your images goes deeper on what happens after an image clears these floors.
Text: character limits are hard, but the business-name check is the sneaky one
Text assets have limits that are enforced exactly: 30 characters on a headline, 90 characters on a long headline, 90 on a description, 25 characters on the business name, and the call to action is either automated by default or picked from a list. Go over and the field simply won't save. The one that catches people isn't the count, it's the business-name rule it has to match your domain or a verified business name and stay free of promotional text and stray symbols. I have seen "Brand 50% Off" style business names rejected not for length but for content, and that rejection can hold up the whole asset group. Note too that in double-width languages like Japanese or Chinese, each character counts as two toward the limit, which quietly halves your usable headline length.
Video: the 10-second minimum is the only real gate
Video is where the specs are loosest and the misunderstanding is biggest. The one hard requirement is duration: a video asset must run 10 seconds or more. There is no video that is technically "required" leave the slot empty and Google auto-generates one from your other assets. Beyond duration the video specs are permissive, but that auto generated clip is not a favour; I would almost always rather supply my own, which brings us to the difference between "valid" and "not limited."
The real minimum to escape "limited" ad strength
An asset group can be perfectly valid every required field filled with a spec-compliant asset and still read "limited." Valid means it will serve. "Good" or "Excellent" ad strength means it has enough variety for the system to assemble competitive combinations across every placement. Those are not the same bar, and the gap between them is exactly where I spend my setup time.
The bare floor to publish an asset group without a Merchant Center feed is modest: three headlines, one long headline, two descriptions, a business name, one landscape image, one square image, and a square logo. That gets you live. It does not get you out of "limited," and a limited asset group is one the system throttles into fewer, weaker placements.
The practical minimum I actually build a performance max asset group to, on every account, is roughly this: eleven or more headlines covering distinct angles, at least one under 15 characters; two to four long headlines; four descriptions; on headlines, descriptions and long headlines I fill well past the required minimum; four-plus landscape images and four-plus squares, with a couple of 4:5 portraits added because they unlock mobile-first inventory the horizontal and square formats can't reach; both the square and the landscape logo; and at least one custom video in each orientation. That set is what reliably pushes an account I run from "limited" to "good" or better. The portrait images and the custom videos are the two slots people skip most and the two that move the rating most, which is not a coincidence they are the formats that open placements the rest of your assets can't.
Ad strength is a completeness-and-variety score, not a performance guarantee; I have run "good"-rated groups that beat "excellent" ones. But a limited group is a real problem, because it is the one state where the spec sheet and delivery are directly linked: the system will not give a starved asset group the inventory it gives a full one. If your groups are stuck there, the fix is almost always more variety, not better individual assets.
What I notice actually moves creative ranking
Once everything is valid and the group is out of "limited," the question shifts from "will it serve" to "which of my assets does the system favour." This is where the reference table is silent and account experience takes over. A few patterns show up consistently enough across the accounts I manage that I plan around them.
Portrait and vertical assets punch above their weight. Every time I have added 4:5 images and 9:16 video to a group that only had landscape and square, the vertical assets started pulling a disproportionate share of impressions because Shorts, Discover and the YouTube feed are vertical-hungry surfaces and most advertisers under-supply them. Supplying the format the auction is short on is one of the most reliable ways I have to earn cheap incremental reach. Custom video beats auto-generated video on almost every account I have compared them on; the auto-generated version fills the slot but rarely earns the ranking a real video does.
The system mixes your headlines descriptions and images per placement, so distinctiveness across those text assets matters more than raw count. Two things I expected to matter turned out not to, at least not the way people assume. Sheer image count past a point of real variety does little twenty near-identical product shots rank no better than six genuinely different ones, and the system seems to reward distinctiveness over volume. And "excellent" ad strength does not translate into better auction performance on its own; it maximises eligibility, which is worth having, but I have stopped chasing the label for its own sake. The place I actually read all of this is the Performance Max asset report, where each asset's Low / Good / Best rating tells you what the system is really favouring that report, not the spec sheet, is where creative ranking becomes visible.
For the raw, authoritative numbers behind everything above every dimension, character count and file-size ceiling Google's official Performance Max specs and format requirements is the reference I keep open while building. Use it for the digits; use everything above for what the digits actually do once the assets are live. And if you want the setting that sits right next to this in my own workflow, the way you resource a group interacts with how it bids my note on how PMax bidding behaves in the learning phase covers the other half of getting a new campaign off the ground.
