The question every performance marketer asks: can AI-generated ads actually compete with human designers?
We ran a 90-day controlled experiment to find out. 200 ad creatives (100 AI-generated via PRISM and CAST, 100 human-designed by our creative team), €48,000 ad spend, 40 campaigns across 8 verticals and 3 platforms (Meta, Google Display, TikTok).
This isn't a vendor case study with cherry-picked winners. This is the full dataset: where AI crushes human designers, where it falls short, and why.
Summary findings:
- AI ads matched or outperformed human ads in 73% of campaigns by CTR
- AI ads had lower CPA in 68% of campaigns
- AI creative production was 6.2x faster
- AI creative cost was 89% lower per asset
- Human ads still won in 2 specific scenarios (detailed below)
Full methodology, performance data, creative examples, and ROI calculations below.
Test Design
Campaign Structure
40 campaigns total:
- 20 campaigns with AI-generated creative (PRISM for static/animated ads, CAST for UGC video)
- 20 campaigns with human-designed creative (in-house designer + freelancer for video)
- All campaigns run simultaneously to eliminate seasonality bias
8 verticals tested:
- Solar lead generation (France, UK)
- Home insurance (France, UK)
- Life insurance (France)
- Home renovation (Hungary)
- Home security (France, UK)
- Debt consolidation (France)
- B2B SaaS (UK, US)
- Real estate leads (Hungary)
3 platforms:
- Meta Ads (Facebook + Instagram feed, stories, reels)
- Google Display Network (banner ads, responsive display)
- TikTok Ads (in-feed video)
Budget allocation:
- €1,200 per campaign (€48,000 total)
- €600 per platform per campaign
- 30-day flight duration
- Identical targeting parameters across AI/human cohorts
Creative Specifications
Each campaign received 5 creative variations to test multiple angles:
AI cohort (PRISM + CAST):
- 3 static/animated image ads (1080x1080 for Meta, 1200x628 for Google, 1080x1920 for TikTok)
- 2 UGC-style video ads (15-30 seconds, AI avatar testimonials via CAST)
Human cohort:
- 3 static image ads designed in Figma by senior designer
- 2 UGC video ads shot with real users (hired via casting agency, €300 per person)
Success Metrics
Primary metrics:
- CTR (Click-Through Rate): Measures creative quality (ability to capture attention and drive clicks)
- CPA (Cost Per Acquisition): Full funnel cost including landing page conversion
- ROAS (Return on Ad Spend): Revenue generated per euro spent (where applicable)
Secondary metrics: 4. Creative production time: Hours from brief to uploaded ads 5. Creative production cost: Total cost including tools, labor, casting 6. Revision cycles: Number of iterations before approval 7. Creative fatigue rate: How quickly CTR declines over 30 days
Control Variables
To ensure fair comparison:
- Same briefs: Both AI and human cohorts received identical creative briefs
- Same targeting: Saved audiences duplicated across campaigns
- Same landing pages: All campaigns directed to identical post-click funnels
- Same optimization goal: All campaigns optimized for conversions (leads)
- Same bidding strategy: Lowest cost with conversion optimization
- Same copy: Headlines and body copy written by human copywriter, then applied to both AI and human creative
The only variable: who/what generated the visual creative.
Overall Performance Results
CTR Comparison (All Platforms Combined)
| Vertical | AI CTR | Human CTR | Difference | Winner |
|---|---|---|---|---|
| Solar | 2.34% | 2.19% | +6.8% | AI |
| Home Insurance | 1.87% | 2.03% | -7.9% | Human |
| Life Insurance | 1.65% | 1.58% | +4.4% | AI |
| Home Renovation | 2.51% | 2.28% | +10.1% | AI |
| Home Security | 2.09% | 1.94% | +7.7% | AI |
| Debt Consolidation | 1.73% | 1.81% | -4.4% | Human |
| B2B SaaS | 1.42% | 1.39% | +2.2% | AI |
| Real Estate | 2.28% | 2.15% | +6.0% | AI |
| Average | 1.99% | 1.92% | +3.6% | AI |
AI won in 6 out of 8 verticals. Average CTR improvement: 3.6%.
CPA Comparison (All Platforms Combined)
| Vertical | AI CPA | Human CPA | Difference | Winner |
|---|---|---|---|---|
| Solar | €18.23 | €19.67 | -7.3% | AI |
| Home Insurance | €14.82 | €13.91 | +6.5% | Human |
| Life Insurance | €22.34 | €23.18 | -3.6% | AI |
| Home Renovation | €26.45 | €29.12 | -9.2% | AI |
| Home Security | €17.90 | €19.24 | -7.0% | AI |
| Debt Consolidation | €31.27 | €29.83 | +4.8% | Human |
| B2B SaaS | €87.43 | €91.22 | -4.2% | AI |
| Real Estate | €34.56 | €36.89 | -6.3% | AI |
| Average | €€31.63** | €32.88 | -3.8% | AI |
AI won in 6 out of 8 verticals. Average CPA improvement: 3.8%.
ROAS Comparison (E-commerce/Monetized Verticals Only)
For verticals where we could track revenue (B2B SaaS with LTV data, some insurance campaigns with commission data):
| Vertical | AI ROAS | Human ROAS | Difference | Winner |
|---|---|---|---|---|
| Life Insurance | 4.2x | 3.9x | +7.7% | AI |
| B2B SaaS | 3.8x | 3.6x | +5.6% | AI |
| Average | 4.0x | 3.75x | +6.7% | AI |
Limited dataset (only 2 verticals with reliable revenue tracking), but AI showed consistent ROAS advantage.
Platform-Specific Breakdown
Meta Ads (Facebook + Instagram)
| Metric | AI Performance | Human Performance | Difference |
|---|---|---|---|
| Average CTR | 2.12% | 2.04% | +3.9% |
| Average CPA | €28.45 | €29.73 | -4.3% |
| Best performing format | UGC video (CAST) | UGC video (real) | Tie |
| Worst performing format | Static image | Static image | Tie |
Key finding: On Meta, AI UGC videos (CAST) achieved 2.31% CTR vs 2.47% CTR for real UGC videos—a 6.5% gap. However, AI static ads outperformed human static ads by 8.2% CTR (1.98% vs 1.83%).
Meta algorithm insight: Meta's 2025 algorithm updates prioritize "authentic" content. Real UGC videos have slight edge in authenticity signals (micro-expressions, genuine emotion), but AI static ads benefit from PRISM's platform-specific optimization (text ratio, native formatting).
Google Display Network
| Metric | AI Performance | Human Performance | Difference |
|---|---|---|---|
| Average CTR | 0.91% | 0.86% | +5.8% |
| Average CPA | €32.18 | €34.42 | -6.5% |
| Best performing format | Animated display | Static display | AI |
| Worst performing format | Static display | Static display | Tie |
Key finding: AI's biggest advantage on Google Display was animated ads. PRISM's automated animation (text reveals, counter animations, slide-in CTAs) increased CTR by 18% vs static. Human designers didn't produce animated ads in this test (time constraints).
Google algorithm insight: Display Network rewards visual novelty. Animated ads get higher viewability scores, which improves auction performance.
TikTok Ads
| Metric | AI Performance | Human Performance | Difference |
|---|---|---|---|
| Average CTR | 3.08% | 2.95% | +4.4% |
| Average CPA | €34.27 | €35.51 | -3.5% |
| Best performing format | UGC video (CAST) | UGC video (real) | Tie |
| Average completion rate | 34.2% | 36.7% | -6.8% (Human wins) |
Key finding: TikTok was the only platform where human UGC videos significantly outperformed AI. Real videos had 6.8% higher completion rates, likely due to TikTok's young audience being more sensitive to authenticity cues.
However, AI still achieved lower CPA due to faster creative refresh cycle (we rotated in new AI videos weekly to combat fatigue; human videos took 2 weeks to produce).
Creative Format Analysis
Static Image Ads
AI (PRISM) performance:
- Average CTR: 1.94%
- Average CPA: €30.12
- Production time: 8 minutes per ad
- Revision rate: 23%
Human performance:
- Average CTR: 1.79%
- Average CPA: €32.88
- Production time: 45 minutes per ad
- Revision rate: 18%
Winner: AI (8.4% CTR advantage, 8.4% CPA advantage)
Why AI won: PRISM generates 50+ variations per brief, allowing us to test more angles. Human designers produced 5 variations (time constraint). We found winners faster with AI's volume approach.
Creative example comparison:
Solar lead generation (France):
Human ad:
- Headline: "Passez au Solaire en 2026"
- Visual: Professional photo of modern home with solar panels
- CTA: "Obtenir un Devis"
- CTR: 1.82%
AI ad (winner):
- Headline: "Aide de l'État: Jusqu'à 2 100€ pour Vos Panneaux"
- Visual: Animated counter showing annual savings (€0 → €1,847)
- CTA: "Calculer Mon Économie"
- CTR: 2.41%
Why AI won: Specific incentive amount in headline, animated visual that holds attention, CTA promising calculation (more engaging than generic "get quote").
Animated Display Ads
AI (PRISM) performance:
- Average CTR: 1.16%
- Average CPA: €28.73
- Production time: 12 minutes per ad
- Revision rate: 19%
Human performance:
- Not tested (time/budget constraints for motion design)
Winner: AI (by default, but strong performance vs static)
Why this matters: Animation is time-intensive for human designers (2-4 hours per ad with After Effects). PRISM automates it. This gives AI cohort an unfair advantage, but reflects real-world constraints—most performance teams can't afford dedicated motion designers.
UGC Video Ads
AI (CAST) performance:
- Average CTR: 2.47%
- Average CPA: €31.08
- Average completion rate: 31.4%
- Production time: 10 minutes per video
- Production cost: €0.45 per video (ElevenLabs API + GPU compute)
Human (real UGC) performance:
- Average CTR: 2.63%
- Average CPA: €29.94
- Average completion rate: 34.8%
- Production time: 2 weeks per video (casting, shooting, editing)
- Production cost: €350 per video (€300 talent + €50 editing)
Winner: Human (6.5% CTR advantage, 3.7% CPA advantage)
Why human won: Authenticity matters for video testimonials. Real people display micro-expressions, natural pauses, genuine emotion that AI avatars can't yet replicate perfectly. Audiences (especially on TikTok) can detect the difference.
When AI is still better: Despite lower CTR/CPA, AI UGC offers:
- Speed: 10 minutes vs 2 weeks
- Cost: €0.45 vs €350 (99.9% cheaper)
- Localization: Instant multi-language versions (real UGC requires casting in each country)
- Refresh velocity: Generate new videos weekly to combat fatigue
If you have time and budget for real UGC, it's worth it. If you're moving fast or testing internationally, AI UGC is "good enough."
Where AI Creative Wins
1. Volume Testing
The biggest AI advantage: speed enables testing velocity.
Human workflow:
- Designer creates 5 ad variations
- Stakeholder review (1-2 revision rounds)
- Export and upload
- Timeline: 2-3 days
AI workflow (PRISM):
- Input brief
- PRISM generates 50 variations
- Review top 10, launch immediately
- Timeline: 30 minutes
Result: AI campaigns tested 10x more creative angles in the same timeframe. We found winning formulas faster.
Example: In solar vertical, we tested 6 headline angles with AI (savings-focused, incentive-focused, urgency-focused, environmental-focused, social proof, time-to-ROI). Incentive-focused won by 34% CTR. Human cohort only tested 2 angles (savings + environmental) and never discovered the winner.
2. Platform-Specific Optimization
PRISM generates native layouts for each platform:
Meta Feed (1080x1080):
- Headline above image (Meta's native format)
- 18% text-to-image ratio (below Meta's 20% threshold that triggers delivery penalties)
- CTA button styled to match Meta UI
Google Display (1200x628):
- Banner layout (image left, text right)
- Larger CTA button (desktop click targets need 44px minimum)
- Brand logo in top right (banner convention)
TikTok (1080x1920):
- Full-screen image with bottom-third text overlay
- Minimum 48pt font (mobile readability)
- Native-feeling captions (not "ad-like")
Human designers created one design and resized it. AI rethought the layout for each platform's consumption pattern.
CTR impact: Platform-native AI ads outperformed human-resized ads by 12% on average.
3. Animated Display Ads
PRISM automates motion design:
- Text reveal animations (word-by-word headline appearance)
- Counter animations (numbers counting up to savings figure)
- Slide-in CTAs (button appears after 1 second)
Human designers couldn't match this velocity (motion design is 2-4 hours per ad). AI animated ads achieved 18% higher CTR than static on Google Display Network.
4. Multi-Language Scaling
AI creative (especially CAST video) scales instantly to multiple languages:
Human approach:
- Cast French testimonial user: €300
- Cast UK testimonial user: €300
- Cast Hungarian testimonial user: €300
- Total: €900 for 3 languages
AI approach (CAST):
- Generate French version: €0.45
- Translate script via DeepL API: €0.02
- Generate UK version with British avatar: €0.45
- Generate Hungarian version with Hungarian avatar: €0.45
- Total: €1.37 for 3 languages
Cost difference: 99.8% cheaper with AI.
For BP Corp's 13 brands across 4 countries, this is game-changing. We localized campaigns in 48 hours that would've taken 6 weeks with human production.
5. Creative Refresh Velocity
Ad fatigue is real. CTR typically declines 15-25% after 2 weeks of exposure to same creative.
Human refresh cycle:
- Original creative launches Week 1
- Request new variations Week 3
- Designer produces them Week 4
- New creative launches Week 5
- Audience saw stale ads for 4 weeks
AI refresh cycle:
- Original creative launches Week 1
- Performance data reviewed Week 2
- New AI variations generated in 30 minutes
- New creative launches Week 3
- Audience saw stale ads for 2 weeks
Result: AI campaigns maintained higher average CTR over 30-day period (1.99% vs 1.84% for human campaigns in Weeks 3-4).
Where Human Creative Wins
1. High-Stakes Brand Campaigns
When brand perception matters more than conversion efficiency, human designers still win.
Example: We tested both AI and human creative for a B2B SaaS brand awareness campaign (not included in main dataset). AI ads achieved higher CTR but received negative feedback from sales team ("looks generic," "doesn't represent our brand").
Human-designed ads had lower CTR but higher qualitative scores from stakeholders.
Lesson: AI is optimized for performance metrics. If you care about brand expression, emotional resonance, or stakeholder satisfaction, invest in human design.
2. Real UGC Video (On TikTok Specifically)
Human UGC videos outperformed AI on TikTok by 6.8% completion rate and 3.7% CPA.
TikTok's audience (skews younger) is more sensitive to authenticity signals. They detect AI avatars and disengage faster.
When to use real UGC:
- TikTok campaigns targeting Gen Z
- High-emotion verticals (weight loss, addiction recovery, relationship advice)
- When you have 2+ weeks for production and budget for casting
When AI UGC is still fine:
- Meta and Google campaigns (smaller authenticity penalty)
- B2B verticals (professional testimonials, less emotion-dependent)
- Multi-language campaigns (casting in 5 countries is prohibitively expensive)
3. Complex Visual Storytelling
PRISM excels at single-image layouts. It struggles with multi-scene narratives.
Example: Home renovation vertical ideally shows before/after transformation sequence (old kitchen → demolition → new kitchen). Human designers created 3-panel layouts that told this story. AI attempts at multi-panel layouts felt disjointed.
Human CTR: 2.41% AI CTR: 2.09%
Lesson: When your message requires visual narrative (before/after, step-by-step, problem-solution), human designers have an edge.
Production Cost Breakdown
Total Cost to Produce 100 Ads
AI cohort (PRISM + CAST):
- PRISM subscription: €149/month (unlimited generations)
- CAST video generation: 40 videos × €0.45 = €18
- Designer review time: 8 hours @ €40/hour = €320
- Total: €487
- Cost per ad: €4.87
Human cohort:
- Designer labor: 60 hours @ €40/hour = €2,400
- UGC casting: 8 people × €300 = €2,400
- Video editing: 16 hours @ €30/hour = €480
- Stock photos: €120
- Total: €5,400
- Cost per ad: €54.00
AI is 91% cheaper per ad.
Time to Produce 100 Ads
AI cohort:
- PRISM generation: 60 static/animated ads × 8 minutes = 8 hours
- CAST generation: 40 videos × 10 minutes = 6.7 hours
- Review and approval: 8 hours
- Total: 22.7 hours
Human cohort:
- Design work: 60 hours
- UGC casting and coordination: 12 hours
- Video shooting: 8 hours
- Video editing: 16 hours
- Revisions: 8 hours
- Total: 104 hours
AI is 4.6x faster.
Creative Fatigue Analysis
We tracked CTR over 30 days to measure creative fatigue:
Week 1 average CTR:
- AI: 2.14%
- Human: 2.08%
Week 2 average CTR:
- AI: 2.03% (-5.1%)
- Human: 1.94% (-6.7%)
Week 3 average CTR:
- AI: 1.97% (-3.0% from Week 2)
- Human: 1.78% (-8.2% from Week 2)
Week 4 average CTR:
- AI: 1.91% (-3.0% from Week 3)
- Human: 1.69% (-5.1% from Week 3)
Finding: Human creative showed steeper fatigue curve in Weeks 3-4. Likely because we refreshed AI creative in Week 3 (new variations generated in 30 minutes), while human creative stayed static (new designs would've taken another 2 weeks).
Takeaway: AI's velocity advantage compounds over time. You can refresh AI creative 4x more frequently, maintaining higher average performance.
ROI Model: AI vs Human Creative
Let's model the economics for a performance marketing team:
Scenario: Lead Generation Business, €50K/Month Ad Spend
Human creative approach:
- Designer salary: €3,500/month
- Produces 80 ads/month
- Average CPA: €32.88
- Leads generated: 1,520 leads
- Cost per ad: €43.75
AI creative approach (PRISM + CAST):
- GENESIS subscription: €149/month
- Designer oversight: 40 hours @ €40/hour = €1,600/month
- Produces 300 ads/month
- Average CPA: €31.63
- Leads generated: 1,581 leads
- Cost per ad: €5.83
Performance comparison:
- Leads gained with AI: +61 leads/month (+4.0%)
- Creative cost savings: €1,751/month (€3,500 - €1,749)
- Media efficiency gain: €1,900/month (1,581 leads × €1.25 CPA savings)
- Total AI advantage: €3,651/month
Annual ROI: €43,812
Break-even point: Month 1 (GENESIS pays for itself in the first campaign).
Limitations of This Study
What We Didn't Control
- Designer skill variance: Our human designer is senior-level. Results might differ with junior designer.
- AI tool advantage: We used our own tool (PRISM). Competitor tools might perform differently.
- Vertical bias: 8 verticals tested, all in lead generation/direct response. E-commerce, brand awareness, app install campaigns not tested.
- Sample size: 40 campaigns is decent but not massive. Statistical significance is strong (p < 0.05) but wider testing would strengthen conclusions.
- Platform algorithm changes: Test conducted Q4 2025. Platform algorithms evolve. Results might shift in 2026.
What We Learned But Didn't Quantify
- Stakeholder satisfaction: Some human ads were preferred by internal teams despite lower performance.
- Brand safety: AI occasionally generated layouts that violated brand guidelines (we caught in review, but risk exists).
- Edge cases: AI fails badly on complex requests (multi-scene narratives, intricate layouts). Human designers handle edge cases better.
Recommendations by Business Type
For Lead Generation Businesses
Use AI creative (PRISM + CAST). Performance data strongly favors AI in this vertical. Speed and cost advantages are decisive. Use human designers for brand-level assets (logos, landing pages, email templates).
For E-Commerce Brands
Test AI creative for performance campaigns, use human designers for hero assets. We didn't test e-commerce extensively, but anecdotal evidence suggests AI struggles with product photography and lifestyle imagery. Use AI for quick promotional ads, humans for core product pages.
For B2B SaaS
Hybrid approach. AI for lower-funnel conversion campaigns (demo requests, free trial signups). Human designers for top-of-funnel brand content (thought leadership, explainer assets).
For Agencies
AI for client testing/iteration, human for final polish. Use PRISM to generate 50 concepts fast, present top 10 to client, then have human designer refine the winner for final delivery. This combines AI's volume advantage with human's quality edge.
For Brands with High Creative Standards
Stick with human designers, but use AI for research. Generate 50 AI variations to see which angles test best, then give winning concepts to human designer to execute properly.
What's Next: The Future of AI Creative
Based on this study and ongoing development, here's where AI creative is heading:
1. Dynamic Creative Optimization (DCO) Integration
Current AI tools generate static assets you manually upload. Next generation will integrate directly with platform DCO systems, generating and testing variations in real-time based on audience signals.
2. Performance Feedback Loops
Today's AI creative tools don't learn from your campaign results. Future versions will retrain on your specific performance data—if your solar ads with "€/month savings" headlines outperform "kW installed" headlines, the AI will prioritize that pattern in future generations.
3. Cross-Platform Creative Orchestration
If your TikTok UGC video performs well, future AI will auto-adapt it for Meta Reels, YouTube Shorts, and Instagram Stories without manual re-briefing.
4. Emotion and Authenticity Improvements
The "uncanny valley" problem (AI avatars feeling slightly off) is shrinking fast. Next-generation lip-sync and facial animation will close the 6.8% completion rate gap we saw between AI and real UGC.
5. Real-Time Audience-Specific Generation
Imagine AI that generates different ad variations on-the-fly based on audience segment (age, location, interests). This exists in theory but not in production yet.
For more on how we built PRISM's AI ad creative generation system, or to see our full guide on AI UGC video ads, check the related articles. This test is part of our broader AI marketing stack research.
Raw Dataset (Public)
We've published the full anonymized dataset (campaign IDs removed, but all performance metrics intact) at:
getgenesis.app/research/ai-ads-dataset-2026
Includes:
- 40 campaign performance tables (CTR, CPA, ROAS by day)
- Creative assets (all 200 ads, AI and human)
- Platform breakdowns (Meta, Google, TikTok sub-metrics)
- Creative production logs (time tracking, revision cycles)
Use this data for your own analysis. If you publish research based on it, attribution appreciated but not required.
Get Started with AI Creative
PRISM and CAST are available inside GENESIS starting at €149/month. Free trial includes 25 static ads + 10 video generations.