Before We Start: A Word on Scepticism
We were sceptical too.
The conventional wisdom in paid social has always been that creative quality is the primary lever — and that quality meant human photographers, professional stylists, and agency art directors who understood composition, emotion, and brand.
That wisdom isn't wrong. But it has an implicit assumption baked in: that "human-made" automatically means "higher quality." That assumption deserves to be tested.
So we tested it.
Over a 90-day period across three Jaiura-powered brands — a women's accessories label, a DTC men's grooming brand, and a home goods company — we ran direct creative A/B tests on Meta (Facebook and Instagram) and TikTok. Same audience. Same copy. Same budget split. The only variable: whether the creative was produced by their agency or generated by Jaiura's AI.
What follows are the seven specific creatives where AI outperformed — the exact setup, the exact results, and the exact prompts used to generate the winning images.
Nothing here is cherry-picked to make AI look infallible. In the full test, human-made creatives won 9 of 23 head-to-heads. We're showing you the 7 where AI won by the largest margin, because those are the most instructive.
How the Tests Were Structured
Before the results: the methodology matters.
- Platform: Meta Ads (Facebook/IG Feed + Stories) and TikTok Ads (In-Feed)
- Budget: Minimum $500 total spend per test. Significance reached at p < 0.05.
- Held Constant: Ad copy, Audience, Bid strategy, Landing page, Time period.
Primary Metrics: CTR (click-through), ROAS (7-day click), and Thumb-stop rate (TikTok only).
Creative #1: The Lifestyle Context Win
Brand: Maren Collective • Product: Crossbody Bag • Platform: Instagram Feed
Professional flat lay: bag on white marble with a succulent and coffee cup. Clean. Neutral. Safe.
Lifestyle: bag on shoulder of woman walking through a tree-lined street at golden hour.
Product: forest green leather crossbody bag
Model: woman, late 20s to mid-30s, autumn street setting
Scene: tree-lined urban street, early evening golden hour lighting
Mood: effortless, put-together but not overdressed
Composition: 3/4 body shot, bag clearly visible at hip level
| Metric | Agency | Jaiura AI |
|---|---|---|
| CTR | 1.8% | 2.7% (+50%) |
| ROAS | 2.4x | 3.6x (+50%) |
Why AI won: The lifestyle context helped shoppers picture ownership. A flat lay shows a product; a lifestyle image sells a version of a life.
Creative #2: The Diverse Model Effect
Brand: Maren Collective • Product: Crossbody Bag • Platform: Facebook Feed
The Agency Creative: Single professional model, studio lit, neutral background. Well-executed standard format.
The AI Creative: Four-panel carousel ad showing the same bag on four different AI-generated models — varying body types, skin tones, and styling. Headline: "Finds its place in every wardrobe."
Why AI won: Reach drove relevance. Showing the same product in four authentic contexts addressed four different customer self-images without a four-day shoot.
Creative #3: The Seasonal Speed Advantage
Brand: Northside Supply Co. • Product: Matte Clay Pomade • Platform: IG Stories
By late January, winter was over. The agency's polished winter assets ($$$) felt dated. Northside used Jaiura to generate spring-context creative in two hours.
CTR: 2.2% (AI) vs 1.4% (Agency) (+57%)
Frequency before fatigue: 4.8 (AI) vs 3.1 (Agency) (+55%)
Why AI won: Creative fatigue was the real opponent. Seasonal agility is a competitive advantage, and AI makes it affordable.
Creative #4: The UGC-Style Challenger
Brand: Northside Supply Co. • Product: Beard Oil Trio • Platform: TikTok
The Agency Creative: $1,200 produced lifestyle video. 28 seconds. Polished.
The AI Creative: A static image designed to look like a lo-fi phone photo screenshot with a text overlay: "Been using this for 3 months..."
Thumb-stop Rate (3s): 51% (AI) vs 34% (Agency) (+50%)
ROAS: 4.1x (AI) vs 2.6x (Agency) (+58%)
Why AI won: TikTok rewards native-feeling content. AI, counterintuitively, can generate "authentic-looking" imagery better than a production team because it avoids the instinct to "over-polish."
Creative #5: The Background Swap Test
Brand: Hazel & Loom • Product: Ceramic Mug Set • Platform: Meta Feed
They kept the product photography identical but used Jaiura to swap backgrounds. Is it the product, or the vibe?
- Variant A (Agency): White surface overhead. ROAS: 1.9x
- Variant B (AI Morning): Timber table, croissant, coffee. ROAS: 2.7x
- Variant C (AI Cosy): Blanket, book, candle. ROAS: 3.4x
Why AI won: Variant C didn't just drive clicks; it increased Average Order Value ($82 vs $68). The creative pre-framed the product as a complete experience.
Creative #6: The Seasonal Colour Variant
Brand: Hazel & Loom • Product: Linen Throw • Platform: IG Feed
They used AI to test product colours (Terracotta, Slate Blue, Forest Green) that didn't even exist in inventory yet.
The Data: Terracotta had a 28% email sign-up rate (vs 12% for the real Cream product). They manufactured 150 units of Terracotta based on this data—and sold out in 11 days.
Creative #7: Direct Response Static vs. Video
Brand: Maren Collective • Product: Spring Collection • Platform: Meta Feed
Agency Video ($3,400): Professional rooftop shoot. ROAS: 3.1x
Jaiura Static Grid ($49/mo): 2x2 grid of four bags in four scenes. ROAS: 3.8x
Why AI won: The 2x2 grid let viewers self-select their favorite product simultaneously. The static performed like a direct-response asset; the video performed like a brand asset.
What the Seven Tests Have in Common
- Context beats production value: Contextual photography sells the experience.
- Authenticity signals outperform polish: Especially on TikTok.
- Speed is a performance lever: AI collapses timelines from weeks to hours.
The Prompt Framework: 7-Element Brief
2. Model: Styling context and age.
3. Scene: Location, time, and season.
4. Mood: 3 adjectives max.
5. Palette: Primary + complementary tones.
6. Composition: Angle and focus depth.
7. Avoid: Crucial negative instructions.
The Bottom Line
AI enables a creative testing velocity that changes the economics of paid social entirely. Most brands test 2–3 variants due to cost. With AI, you test 20. Find winners faster, kill losers sooner.
Survey data and results reflect specific campaign contexts Q1 2025. Agency costs are specific to participating brands and not industry averages.
