Meta Reels AI ad creative specs in 2026 start with one number: 9:16 at 1080 x 1920 pixels, with a safe zone that keeps your hook, logo, and call to action clear of the top 250 pixels and bottom 340 pixels where Meta's UI overlays profile handles, captions, and the comment tray. AI Vidia builds Reels-native creative at that exact spec for DTC and consumer brands across multiple markets, and this guide gives the full format table, the safe zone math, and the AI Vidia protocol for deciding when a Reels-native asset beats a repurposed Feed crop. All figures below come from 1,000+ AI ads and AI stills for our client accounts inside live optimized paid media spend. Get the spec wrong and Meta's delivery system still serves the ad, it just serves it worse: cropped faces, covered captions, and a hook that reads as an afterthought rather than a placement-first creative.
What breaks when Reels specs are wrong
Reels is the highest-volume placement in most Meta ad accounts running in 2026, and it is also the least forgiving on spec. A 4:5 or 1:1 asset dropped into Reels gets pillarboxed with plain bars on either side, which cuts effective screen coverage by roughly a third and drops the product or face size below the point where a thumb-stopping scroll actually stops. Meta's own delivery data continues to reward full-bleed 9:16 creative with lower cost per impression in the auction, because the system treats format-native assets as higher quality signals. The second failure sits inside the frame itself: text, captions, or a logo placed in the top 250 pixels collides with the account handle and follow button, and anything inside the bottom 340 pixels gets covered by the caption line, the like and share icons, and the comment entry field on most devices.
The AI Vidia team treats this as a generation problem, not a cropping problem. Native 9:16 AI video and image generation puts the hook, the product, and the caption-safe copy inside the usable frame from the first render, instead of shooting or generating at 16:9 or 1:1 and hoping the crop survives. That single decision, generate native versus crop later, is the biggest lever most in-house teams are not pulling.
Meta placement specs compared: Reels, Stories, Feed, and in-stream
The table below is the full spec set the AI Vidia team briefs against for every Meta account. Reels and Stories share a ratio but not a safe zone; Feed tolerates more formats but converts worse on mobile scroll; in-stream video keeps the older 16:9 legacy spec for desktop and connected TV delivery.
| Placement | Aspect Ratio | Recommended Resolution | Top / Bottom Safe Zone | Recommended Length |
|---|---|---|---|---|
| Reels | 9:16 | 1080 x 1920 | 250px / 340px | 7 to 15 seconds |
| Stories | 9:16 | 1080 x 1920 | 220px / 250px | 5 to 15 seconds |
| Feed (mobile) | 4:5 or 1:1 | 1080 x 1350 or 1080 x 1080 | No hard overlay zone | 15 to 30 seconds |
| In-stream video | 16:9 | 1920 x 1080 | No hard overlay zone | 6 to 15 seconds |
| Advantage+ shopping | 9:16 primary, 1:1 fallback | 1080 x 1920 | 250px / 340px on 9:16 | 10 to 20 seconds |
Advantage+ shopping placements deserve a specific note. Meta's system will auto-generate a 9:16 version of a 1:1 or 4:5 asset if you do not supply one, using its own cropping and in-painting model, and that auto-crop routinely cuts off a product edge or centers on the wrong subject. Supplying a native 9:16 version yourself removes that auto-crop from the equation entirely and is one of the fastest wins available to a media buyer this quarter.
The AI Vidia Reels Placement Fit Protocol
This is the strategic decision model AI Vidia uses to decide, brand by brand, how much of the creative budget goes to Reels-native production versus repurposed Feed and Stories assets.
- Audit the current aspect mix. Pull the last 90 days of ad spend by placement and cross-reference against asset aspect ratio. Most accounts the AI Vidia team inherits are running 60 to 80% of Reels spend on a repurposed 1:1 or 4:5 asset, which is the single largest fixable leak in the account.
- Score each creative concept for native fit. A concept built around a tight product macro or a face-forward UGC frame translates well to 9:16 from the first render. A concept built around a wide lifestyle scene or a group shot loses too much information when generated vertically and should stay in Feed and Stories rather than being forced into Reels.
- Segment budget by fatigue window. Reels creative fatigues in 5 to 7 days at the spend levels the AI Vidia team sees across our client accounts, roughly half the fatigue window of Feed static. Placement-level budget needs a matching placement-level refresh cadence, not one shared creative calendar.
- Generate native at capture, never crop after. Set the AI generation brief to 9:16 from the first prompt for any concept scored for Reels. Cropping a 16:9 or 1:1 render after generation reintroduces the safe zone problem the whole protocol exists to avoid.
- Gate every asset on AI disclosure compliance. Meta's automated classifiers throttle delivery on unlabelled synthetic depictions of real people, places, or events within roughly 48 hours of detection. Build the disclosure check into the QA gate, not as an afterthought before launch.
- Assign hook-testing volume by placement, not by account. Reels earns 20 to 40 new variants per week in a healthy account; Feed and in-stream can run on 8 to 12. Splitting variant volume evenly across placements under-serves Reels and over-produces for slower-fatiguing formats.
The practical shift this implies: brief the AI generation stack in 9:16 first, then generate 4:5 and 1:1 variants as separate renders with their own safe framing, rather than asking one master asset to serve every placement through post-hoc cropping.
