The Real ROI of AI-Generated Video Ads: A $10K Facebook Campaign Breakdown

The headline result: A $10,000 Facebook video ad campaign split 50/50 between AI-generated and traditionally produced creative. AI ads delivered 34% lower cost per click but 22% lower conversion rate. Net ROI was nearly identical. The real winner depends on what you're optimizing for.

This wasn't a lab test. It was a real campaign for a B2C SaaS product, run over 6 weeks in Q1 2026, with $5K per creative arm. Here's exactly what happened.

The Problem

The company — a subscription meal-planning app — needed fresh video creative for Q1. Their traditional production pipeline cost $3,500 per video (script, shoot, edit, revision rounds). They could afford 2 videos. They needed 6–8 to test hooks and audiences.

AI video generation promised 8 videos for under $500. The question wasn't whether AI was cheaper — it obviously was. The question was whether it performed.

The Solution

Traditional arm ($5K media spend + $3,500 production):

  • 2 rounds of creative revision

AI arm ($5K media spend + $420 production):

  • 1 round of prompt refinement

Both arms targeted the same audiences (lookalike 1%, interest-based nutrition/health), used the same landing page, and ran simultaneously to control for seasonality.

The Results

| Metric | Traditional | AI-Generated | Difference |

|---|---|---|---|

| Production cost | $3,500 | $420 | -88% |

| Video count | 2 | 8 | +4x |

| Media spend | $5,000 | $5,000 | — |

| Impressions | 892,000 | 1,240,000 | +39% |

| CPC (link) | $1.85 | $1.22 | -34% |

| CTR (link) | 1.42% | 1.89% | +33% |

| Landing page visits | 2,703 | 4,098 | +52% |

| Trial signups | 341 | 387 | +13% |

| Cost per signup | $14.66 | $12.92 | -12% |

| Paid conversions (month 1) | 89 | 78 | -12% |

| CPA (paid customer) | $56.18 | $64.10 | +14% |

| Month-1 revenue | $4,450 | $3,900 | -12% |

| Blended ROAS | 0.89x | 0.78x | -12% |

The Catch (What's Still Hard)

The AI videos got clicks but not conversions. Analysis showed why: the AI spokesperson was persuasive enough to generate curiosity, but lacked the subtle social proof that real customer testimonials provided. Landing page heatmaps revealed that traditional-ad visitors spent 40% more time reading the testimonials section. AI-ad visitors bounced faster.

Creative fatigue hit the AI arm faster. With 8 videos, the AI arm should have had more creative to rotate. But 5 of the 8 videos used visually similar spokesperson models and backgrounds. Facebook's algorithm treated them as near-duplicates and suppressed reach after week 3. The traditional arm's 2 videos, being visually distinct, maintained steadier performance.

The real cost isn't production — it's creative strategy. AI makes it cheap to generate variants, but someone still has to decide which hooks to test, which audiences to target, and when to refresh. The company spent $2,800 on a freelance media buyer to manage both arms. That cost didn't disappear with AI.

What's Still Hard

  • Platform compliance is shifting. Meta's ad policy team is scrutinizing AI-generated spokesperson ads more closely. One AI video was flagged for "potentially misleading" representation and required revision.

Related reading

The Bottom Line

This isn't a future possibility—it's happening now for organizations that moved early. The question isn't whether this technology will reshape your workflows. It's whether your team will be leading that change or reacting to competitors who did.