Barry Hott's Method: Making AI Native Ads Look Human
A Facebook ads veteran demonstrates how Google's VO3 can now generate human avatars so realistic that audiences can't tell them apart from real footage—then explains why the "make ugly ads" philosophy matters more than ever as AI floods social feeds with perfect-looking content.
Read Original Summary used for search
TLDR
• AI video generation (VO3) has crossed the uncanny valley—Barry shows insurance ad examples where most viewers can't identify which person is AI vs. real, including one that's entirely AI-generated with natural voice inflection
• The "ugly ads" philosophy: your competition isn't other ads, it's everything in the scroll—Mr. Beast deliberately uses cheaper cameras because audiences don't want overly polished content
• Authenticity arms race: as AI gets better at mimicking human content, real creators need to do things AI can't replicate yet (imperfections, shaky cam, unique moments that wouldn't be in a prompt)
• Speed beats perfection: teams waste time tweaking ads pre-launch instead of learning from live performance—the faster you iterate, the faster you learn
• Study what your actual audience consumes by creating separate Instagram accounts for different interests, then make content that matches those aesthetics
In Detail
Barry Hott, who's studied billions in ad spend over 17 years in Facebook ads, makes a case that fundamentally challenges how marketers think about creative. His core thesis: audiences aren't on social to see ads—they're consuming hyper-personalized organic content, so your ads must blend in by matching what people actually watch, not what marketing textbooks say looks "good." This often means deliberately making content that looks less polished, shot on phones rather than professional cameras, because that's what performs. He cites Mr. Beast, who has infinite resources but deliberately stopped using 4K/8K cameras because audiences didn't want overly polished content.
The presentation takes a darker turn when Barry demonstrates Google's VO3 video generation. He shows multiple examples of insurance ads, asking the audience to identify which are AI vs. real. Most get it wrong. The final reveal is genuinely unsettling: a woman walking and talking naturally about insurance rates is entirely AI-generated, including her voice with natural inflection and pacing—created from just a text prompt describing her appearance and the script. He shows his actual VO3 workflow, demonstrating how he can generate multiple realistic human avatars in minutes, something that would have required finding the right creator, scheduling shoots, and multiple takes.
This creates what Barry calls the "authenticity arms race." As AI gets better at mimicking human content (and it's improving daily), real creators need to do things AI can't or wouldn't do—imperfections, shaky camera work, unique moments that wouldn't naturally appear in a prompt. But there's a paradox: AI content can be "too perfect" (like the smooth falcon footage vs. the real shaky version), while traditional marketing content is also too perfect. The winning content exists in a sweet spot of authentic imperfection. Barry's tactical advice: stop tweaking ads before launch and start learning from live performance faster. Create separate Instagram accounts to study what content actually performs in different niches. Use AI as a tool to support storytelling (showing problems, creating scenarios you can't afford to shoot), but prioritize building authentic authority now, before AI makes everything questionable. The window to differentiate with human authenticity is closing.