Static ads for Facebook

Facebook static ad specs

Meta's official specs, with the ones that actually matter in bold.

File type: JPG or PNG. Keep under 5 MB in practice. Anything larger will be auto-compressed by Meta and you lose control over the final look. Text limit on primary text: 125 characters before truncation. Meta allows up to 3,000 but the fold happens at 125.

Ad placements: where static works

Static creatives run across every Meta placement, which is one of the reasons static beats video on reach economics. Video placements accept static, static-only placements reject video. Full coverage costs one creative.

• Feed (Facebook + Instagram). Highest volume placement. 4:5 mobile-native outperforms 1:1 by 15 to 25% on conversion objectives.
• Marketplace. Under-rated for e-commerce. Users are already in a buying mindset. Static product-in-context ads perform well here.
• Right Column. Desktop-only, 1.91:1 banner. Cheap CPMs, mostly retargeting.
• Stories + Reels. 9:16 full- screen. Static works, but the creative language is different: bigger typography, safe zones for platform UI at top and bottom (~250px each).
• Explore + Search. Instagram surfaces. Static feed-native creatives are accepted.

The Meta text-in-image penalty: myth or real?

The formal 20% text rule ended in 2020. Meta publicly said the rule was over. So is text-heavy static still penalised? The honest answer is: informally, yes.

Meta's auction rewards ads that users engage with. Users engage less with wall-of-text ads. The algorithm learns this and delivers your text-heavy ad to fewer impressions at higher CPMs. There is no rejection, no rule violation. Just quieter delivery.

The practical rule: one hook (5 to 9 words), one supporting line (5 to 12 words), one CTA (2 to 4 words). Any ad that fits on one line at thumbnail size passes. Any ad that doesn't is competing with itself.

What Meta's algorithm actually rewards

Three signals, in order of importance for the auction:

1. Thumbstop rate. The percentage of users who stop scrolling on your ad. Static wins or loses in the first 400 ms. Product-in-context or a real face beats stock illustration every time.
2. Engagement quality. Comments, saves, and shares matter more than reactions. Reactions are cheap. Saves signal purchase intent.
3. Downstream conversion. The click plus what happens after. Meta's Advantage+ campaigns pass this signal back into ad delivery, which is why post-click UX matters as much as the ad.

The algorithm does not reward polish. It rewards the outcome. A rough-looking static that closes at a lower CPA gets more delivery than a beautiful one that doesn't.

7 static ad patterns that win on Meta in 2026

Patterns pulled from our library of 800+ winning ads:

1. Product close-up. Detail-shot of the product with a specific benefit hook overlaid. Works for physical goods, apparel, beauty, consumer electronics.
2. Lifestyle-in-context. Product being used in the wild, real environment, real person. Higher trust signal than studio photography.
3. Testimonial card. Customer quote in large type, name and photo underneath. Works especially for higher-price categories where proof unlocks the click.
4. Comparison. Two options side by side, ours vs the alternative. Direct, punchy, works when your product beats a specific competitor category.
5. Before/After. Split image showing the transformation. Works for fitness, skincare, home renovation, weight loss, financial products.
6. Urgency banner. Sale ends Sunday, last 24 hours, limited stock. Only works when the urgency is real. Fake urgency destroys trust.
7. Problem-solution. Named problem in the hook, named solution in the visual. Works for pain-point-heavy categories: back pain, sleep, productivity, meal planning.

Common mistakes

• Wrong format for placement. Uploading a 1:1 for Stories. Meta will crop or letterbox it and your creative dies.
• Too much text. Users cannot read it at thumbnail size, and the algorithm quietly throttles delivery.
• Low contrast. Grey text on a busy photo background. The hook has to punch through the visual, not fade into it.
• Stock imagery. Users recognise stock in a fraction of a second. Trust drops, click rate drops with it.
• Only 3 creatives per campaign. Under 10 ads is not enough variance for the algorithm to find winners. Budget goes to whichever ad launched with the best first-hour engagement, not the ad with the highest ceiling.

Meta ad account structure: where static fits

For most performance accounts the recommended structure in 2026 is:

• Campaign level. One Advantage+ campaign per conversion objective. Let Meta optimise placements.
• Ad set level. Broad targeting. Meta's targeting has been eroded by iOS 14+ and Advantage+ works better with wide audiences.
• Ad level. 30 statics per ad set, across 5 angles. Refresh top 5, kill bottom 25 every 5 to 7 days.

If you're running TikTok in parallel, the same batch of 30 statics adapts to 9:16 for that platform. See static ads for TikTok for platform-specific specs. For the step-by-step production flow, see how to make static ads. For a template library, see static ads templates.

FAQ

What size should a Facebook static ad be?

1:1 (1080x1080) for feed, 4:5 (1080x1350) for mobile-first Instagram and Facebook feed, 9:16 (1080x1920) for Stories and Reels. Build all three natively rather than letting Meta crop.

Does Facebook still penalise too much text?

The formal rule ended in 2020, but text-heavy ads still perform worse in practice because engagement drops and the auction learns from it. Keep it to hook plus supporting line plus CTA.

What placements accept static?

All of them. Feed, Marketplace, Right Column, Stories, Reels, Audience Network, Messenger. Video placements accept static too, so static gives you full coverage on one creative.

What wins on Meta in 2026?

Product-in-context or real faces plus buyer-voice hooks plus batches of 30 for algorithmic variance. Polish comes second. Outcome comes first.

How much to spend testing?

$50/day per ad set for 3 to 5 days. That's $150 to $250 to learn if a batch of 30 has winners. Below that spend, you're reading noise.