AI Image Generation for Ad Creative: What Actually Works vs. Hype

Three months ago, a home services client asked if they should start using AI-generated images in their Google Ads instead of hiring a photographer. They had a $3,000/month budget and spent $400 on monthly photoshoots. The math felt obvious—use AI, save the money. We tested it. Generated 50 images of technicians, before-and-afters, and happy families using Midjourney and DALL-E 3. Ran them against their existing ads. The results surprised us: AI images underperformed by 31% on click-through rate and 18% on conversion rate. But there was a specific use case where AI crushed it. Here's what we learned.

Where AI Images Fail: Trust-Critical Moments

In home services, HVAC, plumbing, and roofing—people are inviting someone into their home to do $2,000-15,000 of work. They need trust. AI-generated images of technicians, while improving rapidly, still read as slightly "off" to the human brain. We tested an AI image of a technician in a blue shirt holding tools next to a real photo of the same scenario. Real photo: 3.2% conversion rate. AI image: 2.6% conversion rate. The difference was enough that the real photo was worth the $400/month investment.

Local service businesses live and die by credibility. When someone searches 'emergency plumber near me' at 11 PM with a pipe burst, they want proof that a real, competent person will show up. Real images of actual customer work, testimonials with real faces, and process photos of your actual team beat AI every time in this category. Don't replace them.

Where AI Images Win: Process, Concepts, and Lifestyle

Here's where it flipped for us: conceptual images. The same client ran ads highlighting benefits like 'Emergency Response,' 'Preventive Maintenance,' and 'Peace of Mind.' These don't need a specific technician or customer—they need a vibe. We generated 12 variations of 'professional HVAC team in a modern dispatch center,' 'family relaxing in comfortable home,' and 'technician reviewing maintenance checklist.' These ads had comparable performance to stock photos but at zero cost. CTR was 2.1% vs. stock photo's 2.0%. Conversion rate: 2.4% vs. 2.3%. Statistically equivalent, but AI cost $12 to generate, stock photos cost $25-50 each.

We found the sweet spot with a plumbing company: use real photos for 70% of ads (actual work, actual team, testimonials), and use AI-generated conceptual images for 30% (lifestyle, benefit-focused, seasonal themes). This hybrid approach meant they could scale creative volume—test 5 variations on 'winter pipe protection' for $36 instead of hiring a photographer—without sacrificing trust on the hero creative.

• Use AI for lifestyle, conceptual, and benefit-focused imagery (not before-and-afters)
• Keep real photos for trust-critical moments: your team, past work, customer testimonials
• Test AI images against your existing best performers—don't assume they'll work
• Generate multiple variations quickly and test them (batch of 20 variations costs $15-30)
• Watch for the 'uncanny valley' tells—odd hands, asymmetrical faces, strange backgrounds—users notice

The Real ROI: Speed and Volume, Not Cost Savings

Stop thinking about AI images as a way to eliminate your photography budget. Think about them as a way to test 10x more creative variations in the same timeframe. A local salon spent $200/month on professional photos. We generated 50 AI images of different hairstyles, color treatments, and salon environments for $30 total. We ran them as test ads for a week. Three variations crushed the existing photos by 40%+ CTR. The salon then hired a photographer to shoot those exact scenarios in real life, which became their new hero creative. AI helped us identify winning concepts in days instead of weeks of iteration.

This is the real value: speed to optimization. In paid advertising, the first iteration is rarely your best. You need to test 20-50 variations to find winners. AI lets you generate those variations in hours, not weeks. A fitness studio ran 40 different gym lifestyle ads (AI-generated) for a month at $5/day total budget. The top 3 performed 35% better than their baseline. They then allocated their $2,000/month ad budget to scale those winners with real photos. Without AI testing, they would've taken 6 months of trial-and-error to find those winners.

AI images aren't about replacing photographers. They're about testing 10 creative directions in a week for $30 instead of spending $2,000 on a shoot that might miss the mark.

← Back to all articles