GOOGLE ADS DEMAND GEN SPECS: THE CREATIVE RULES THAT ACTUALLY MATTER

The thing nobody tells you about google ads demand gen specs is that the table Google publishes is the easy part the hard part is which assets the system quietly refuses to serve even after it accepts the upload. I've had ads sit at "Approved" while a third of the placements they were built for never received a single impression, because one aspect ratio was missing or one logo was still under review. This is the view from someone managing live budgets, not a copy of the spec sheet: the limits that matter are the ones that decide whether your creative ranks, gets clipped, or never competes for the inventory you paid to reach.

So I'll skip the wall of numbers you can read on Google's page and focus on what I actually watch. There are five things the demand gen google ads specs hinge on in practice: image aspect ratios and file size, the square logo and its crop, the headline and description character limits, the video rules, and the way full coverage feeds creative ranking. If you want the platform background first, the primer on what Demand Gen is sets the stage; this piece is the asset layer specifically, and where it bites.

One thing worth setting straight before the specs: this is a single campaign type that serves the same asset set across very different surfaces. Google demand gen campaigns run across YouTube, the Gmail and Discover feeds, the Google Display Network, and YouTube Shorts, and each of those renders your creative differently. That's the whole reason the specs feel fussy a square image that looks clean in a Gmail promotions slot gets letterboxed in a full-screen Shorts placement, and a landscape video that owns the in-feed unit can't fill a vertical one. The specs aren't bureaucracy; they're Google's way of making sure you've supplied something usable for every place the ad can show. When I read the table, I'm really asking one question per row: which placement does this asset unlock, and what does my reach lose if I skip it.

Image specs: the four ratios and the one that gets rejected

The image side of the google ads demand gen specs comes down to four aspect ratios: horizontal 1.91:1 (I export at 1200x628), square 1:1 (1200x1200), and two verticals 4:5 (960x1200) and 9:16 (1080x1920). Each uploaded image caps at 5 MB and you can attach up to 20 per ad. Those are the numbers. The inference that matters is which ones the system rejects: the 9:16 image, the newest format and the one built for Shorts, can be refused outright when the campaign runs through Display & Video 360, because it isn't supported there. If a single image keeps bouncing on a DV360 setup, that's almost always why and the fix is to lean on 4:5, 1.91:1 and 1:1 for that inventory rather than fighting the rejection.

The lever underneath all of this is the rule of three. Google's own guidance is at least three images per aspect ratio per ad group, and it isn't a nicety an ad group with one square and one landscape image doesn't get rejected, it just quietly fails to reach most of the inventory, because the system has nothing to rotate into the placements it can't fill. I treat the missing 9:16 the same way I treat a missing asset anywhere: not a compliance gap but a reach gap. Feed it three strong variants per ratio and it competes everywhere; feed it the minimum and you've capped your own delivery before the auction even runs.

Where this gets practical is in how you structure your ad groups. I keep asset sets tight and themed per ad group rather than dumping every image into one, because the rule of three applies per ad group two focused groups each fully covered beat one sprawling group with a few images per ratio. The vertical formats are where I'm most disciplined: the 4:5 and especially the 9:16 are the assets people skip, and they're exactly the ones that unlock the mobile-first, full-screen inventory that carries a lot of Demand Gen's volume. So the inference I'd pass on is simple: build the verticals first, not last, because they're the ones the rest of the format set can't substitute for.

The logo is the asset I see rejected most

If one asset in the demand gen google ads specs causes more disapprovals than the rest combined, it's the square logo. It has to be 1:1, minimum 128x128 but I always supply 1200x1200, and here's the part the spec sheet states but people skip: Google crops roughly 21% of it into a circle on Gmail about 5% off each of the four corners. Anything designed edge to edge gets clipped, and a clipped or off-spec logo is one of the most common reasons an otherwise fine ad gets disapproved or just renders broken. Centre the mark, give it real margin, and assume the corners will vanish.

The second logo trap is timing, and it's caught me on launches. Logo review can take five to seven working days, and a logo that's still under review is not the same as an approved logo the ad can't serve against an asset that hasn't cleared. So when I'm spinning up a campaign on a brand-new logo, I upload it first and build everything else while it clears, rather than discovering on launch day that the whole thing is held. One more wrinkle worth knowing: the logo file-size ceiling shifts by format 5120 KB on the single image ad but only 150 KB on carousel and video so a logo that passed on one ad type can fail on another for size alone.

Headline, description and CTA: the character limits that bite

The text limits are blunt and the editor enforces them at the ceiling, so the question isn't whether you'll exceed them you can't it's how you write inside them. Headlines run to 40 characters, up to five of them. Descriptions are 90 characters, again up to five. Business name is 25 characters and is required. On video ads there's an extra long headline at 90 characters for the in-feed format, and the call to action is capped hard at 10 characters, which is tighter than people expect and where I most often have to cut a word.

The inference that changed how I write these: on YouTube Shorts the system treats headline, long headline and description as interchangeable and can use any of them as the ad's description. That means the old habit of writing one hero headline and treating the rest as filler quietly costs you, because any of the five might be the line a user actually sees. So I write all five headlines as if each has to stand alone same discipline I bring to Demand Gen best practices generally, where the assets you treat as afterthoughts are the ones the system serves when you're not watching. Headline character limit isn't just a count; it's five chances, and you want all five working.

Video specs and the 10-second cliff

Video in Demand Gen lives on YouTube you reference it by URL, you can't drop an offline file straight into the ad and you can attach one to five videos per ad. The supported ratios are 16:9, 1:1, 4:5 and 9:16, and I supply at least 16:9 and 9:16 so one campaign covers in-feed and Shorts without the system having to centre-crop a landscape video into a vertical slot, which mangles the subject every time. Length is where the real gate sits: the floor is five seconds, but anything under 10 seconds won't serve on YouTube in-stream at all, and Google recommends 15 seconds or more for reach. That 10-second cliff is the one people trip over, building a snappy six-second cut and wondering why in-stream stays empty.

One more piece of the google ads demand gen specs worth flagging: if you don't supply a video, Google can auto-generate one from your images and final URL. It's a real fallback, but in my experience the auto-built videos underperform a proper upload often enough that I treat custom video as the default and the generated one as an emergency, not a plan. The same logic runs through how I think about Demand Gen campaign setup overall automation fills gaps competently, but the assets you build deliberately are the ones that carry the campaign.

If I had to name the single video asset that earns its production cost, it's vertical video at 9:16. That's the format that serves the high-attention placements where Demand Gen leans heaviest the youtube shorts discover and in-feed surfaces and a centre-cropped landscape clip is not a substitute it crops the subject out of frame and reads as a repurposed afterthought, which is exactly what it is. Two practical notes from live accounts: keep the action and any text inside the safe zone, because Shorts overlays UI on the edges, and keep captions on, since a large share of vertical impressions render with the sound off by default. Get the vertical built properly and you cover the highest-attention inventory; lean on a cropped horizontal and you'll watch that placement quietly under-deliver while the dashboard still says the ad is "serving."

Product feeds, targeting and where the specs meet reporting

The specs don't stop at images and video. If you're a retailer, Demand Gen can pull a product feed straight into the ad, and that feed has its own requirements that sit alongside the creative ones. The feed lives in your Merchant Center account, and the product images, titles and prices it serves follow Shopping's feed rules, not the single-image spec so a feed that's disapproved or thin in Merchant Center starves the ad regardless of how clean your uploaded creative is. I check the feed health first when a product-feed ad under-delivers, because the spec problem is often upstream of the campaign entirely. One eligibility quirk worth knowing: the product-feed-only format isn't eligible on YouTube in-stream, so if you're feed-only and expecting in-stream reach, you won't get it until you add image or video assets.

Audience targeting interacts with the specs in a way that's easy to miss too. The asset coverage you supply decides which placements you can serve on, but your audience targeting decides who sees them there and a mismatch between the two wastes both. There's no point building a perfect 9:16 pack for Shorts if your targeting is pushing the campaign toward Gmail-heavy audiences, and vice versa. I line the two up deliberately: the formats I invest in should match the surfaces my audience actually lives on, rather than building every asset to a uniform standard and hoping the system sorts it out.

And the place all of this becomes visible is the asset report. Demand Gen reports performance at the ad group level and, within that, per asset so the way you split your ad groups directly shapes how readable your data is. Cram mismatched themes into one ad group and the ad group level numbers blur together. I keep groups clean specifically so the report stays legible: each asset gets a Low, Good or Best label based on served performance, and I rotate the Low-rated ones out every couple of weeks. Read at ad group level, swap the weakest assets, keep coverage full that loop is what turns the specs from a one-time checklist into ongoing optimisation.

What I notice actually moves creative ranking

Here's the distinction that took me too long to internalise: meeting the specs gets you approved, but covering them fully is what lets the ad rank. Demand Gen can only rank what it can serve, so a thin asset pack that technically passes every limit still loses a missing 9:16 means the ad never competes for Shorts, a missing square video means it skips a chunk of feed inventory, and no amount of bid or budget buys back placements the creative can't physically fill. The asset coverage is the floor of your reach, not a checkbox you tick once.

The signal I watch for this is ad strength. An ad group with every aspect ratio filled and three-plus assets each consistently shows stronger ad strength, and stronger ad strength means the system has the room to push budget toward the combinations that actually convert. I read it the way I read the diagnostics in any Demand Gen frequency cap review: not as a grade to chase for its own sake, but as a proxy for how much inventory the creative can compete for. Full coverage, varied assets, all five headlines pulling their weight that's what I've seen move ranking, far more than any single "best" image or clever line.

The practical takeaway

If you remember one thing about google ads demand gen specs: the published table tells you what's allowed, but it won't tell you what gets quietly refused or under-served. Cover all four image ratios with three variants each, centre the square logo with margin and upload it early so it clears review, respect the 10-second in-stream floor on video, and write all five headlines as standalone lines because Shorts may use any of them. Those are the rules I actually check before a Demand Gen ad goes live, in that order.

So I no longer treat the spec sheet as a one-time compliance pass. I treat it as a coverage map every ratio I skip and every asset I under-supply is inventory I've handed back before the auction starts. Get the coverage right and the system has room to rank you; get it thin and it faithfully serves your gaps. The full, current numbers live in Google's official Demand Gen asset specifications read that for the exact figures, then read this for what they cost you when you cut corners.