Two problems account for most of the creative rework in a performance team: text that gets covered by the platform's own interface, and the sheer volume of resizing needed to cover every placement. They are usually treated as separate tasks. They are the same problem, and solving them together is what turns a two-day production job into a single pass.
What a safe zone actually is
Every vertical ad is displayed underneath the platform's interface. Profile bars, captions, CTA buttons, icon rails — none of them are part of your creative, all of them are drawn on top of it, and none of them appear in your design tool. The region guaranteed to stay visible is the safe zone, and it differs on every platform.
This is why a creative can pass every internal review and still ship with its call to action invisible. Nothing was wrong with the design; it was evaluated in the wrong context.
The overlay maps (1080 × 1920 canvas)
| Platform | Top overlay | Bottom overlay | Side overlay |
|---|---|---|---|
| Instagram Stories | ~250 px | ~340 px | minimal |
| Instagram / Facebook Reels | ~180 px | ~420 px | right ~120 px |
| TikTok | ~130 px | ~500 px | right ~140 px |
| YouTube Shorts | ~120 px | ~400 px | right ~140 px |
Horizontal placements have overlays too, though smaller: YouTube thumbnails carry a duration badge over the bottom-right corner, and Google PMax draws its own headline and CTA adjacent to your image rather than on it.
The universal safe window
Intersect the four vertical maps and you get the region that survives everywhere: roughly x: 40–940, y: 250–1420 on a 1080 × 1920 canvas — about 55% of the frame, centred slightly above the vertical midpoint.
Designing your subject, headline and logo inside that window means one 9:16 export works on all four platforms with no per-platform variants. You give up some canvas, and in exchange you delete an entire category of production work and an entire category of shipping error. For most teams that is a clearly favourable trade.
Why centre-cropping breaks safe zones
Safe-zone failures rarely come from bad design. They come from good designs being cropped.
A 1:1 feed creative correctly places its text near the top and bottom edges — that is where it belongs in a feed context. Centre-crop that same creative to 9:16 and those text blocks land precisely inside the Stories overlays. The design was right, the conversion was wrong, and no amount of reviewing the original would have caught it.
The three ways teams handle multi-ratio production
| Approach | Speed | Design fidelity | Failure mode |
|---|---|---|---|
| Manual re-layout per ratio | Slow (30–60 min per set) | Perfect | Does not scale; the queue becomes the bottleneck |
| Centre crop / platform auto-crop | Instant | Poor | Amputates subjects, breaks safe zones |
| Generative "extend / reimagine" | Fast | Unpredictable | Invents elements that were never approved |
Why generative reimagining is a trap for ads
This deserves separating from the other two, because it is the option that looks most modern and carries the least obvious risk.
An ad creative is an approved artifact. Brand, legal and often regulatory review signed off on a specific image. A generative fill that adds a hand, changes a product angle or invents background objects produces a technically attractive image that is no longer the approved creative. For regulated categories that is a compliance incident, not a design preference.
For performance teams the risk is subtler but just as real: hallucinated elements shift what the ad communicates, so differences in performance between ratios can no longer be attributed to the variables you were actually testing. You lose the ability to read your own results.
The preservation-first workflow
A resize that respects the original design does four things, in order:
1. Understand the source. Identify the subject, text blocks, logo and background regions — what the design is — before touching a pixel. 2. Re-solve the layout per ratio. Each target canvas gets its own arrangement of the same elements, with that platform's overlay map subtracted from the usable area. 3. Extend backgrounds from source material only. Where the new canvas needs more background, continue it from the image's own texture rather than inventing content. 4. Verify against the original. Same elements in, same elements out — nothing missing, nothing added, text unchanged.
A worked example
Below is a real AutoAdfit export running that workflow on an extreme ratio change — a 3:1 banner reformatted to a 1:1 square. The change in canvas shape is far larger than a typical feed-to-story conversion, which makes the behaviour easy to inspect.
SOURCE · 1200×400
OUTPUT · 1024×1024
The second example applies the same process in the opposite direction — a square creative with content anchored near both side edges, reformatted to a wide 16:9 canvas where a centre crop would have removed the edge elements entirely.
SOURCE · 1080×1080
OUTPUT · 1376×768
A practical QA checklist
Whatever tool produces your set, check each ratio against four questions before trafficking:
Is every element from the original present? Is anything present that was not in the original? Is all text outside that platform's overlay zones? Does the subject still read at thumbnail scale? A creative that passes all four on every ratio is ready to ship, and one that fails any of them will cost more to diagnose after launch than to fix now.
Safe-zone-aware formats in AutoAdfit
Every 9:16 export below is generated with the corresponding platform's overlay map applied:
| Platform | Format | Size |
|---|---|---|
| Google PMax | Story | 1080 × 1920 px |
| Meta Ads | Story | 1080 × 1920 px |
| YouTube | Shorts | 1080 × 1920 px |
| TikTok | Video | 1080 × 1920 px |