Field notes on AI ads.
Original research, opinions, and tactics on AI in paid advertising — from the team building AdControlCenter.
Hands-on: building a logo overlay system that doesn't look fake
Most AI-placed logos look pasted on because they ignore what's already in the image — here's the two-step vision pipeline we built to fix that.
The evaluation framework that wins, not the model
Switching AI models gives you a marginal lift. Building a real evaluation loop gives you the compounding advantage.
Product fidelity in AI ads: catching when the model swaps your product
AI image models will silently replace your actual product with a plausible-looking substitute — here's exactly how to detect and prevent it.
Direct Response advertorial images: a working 6-step example
How to go from a blank canvas to a conversion-ready advertorial image for supplements—using ChatGPT, Gemini, and one spy tool—without touching Photoshop.
The G.E.M framework, but for static ads
The G.E.M framework was built for AI video—but its Generate / Extract / Multiply logic turns out to be exactly what broken static ad workflows need.
The 2-edit rule: why your AI ad gets worse the more you tweak it
Every additional edit you make to an AI-generated ad image compounds the model's errors — here's the exact mechanism, and what to do instead.
JSON prompts vs prose prompts: a 50-ad side-by-side test
We ran 50 ad image prompts through Ideogram v3 in both prose and JSON formats — the format that wins depends entirely on the category, and the loser surprised us.
9 ways to multiply one product image into a full ad set
One product photo is enough raw material for a complete, platform-ready ad set — here are the exact nine templates we use to get there.
How AI Image Generation Is Changing Ad Creative
Three years ago a single ad image cost $200 and a week. Today it costs ten cents and ten seconds. The economics of testing changed completely — but most operators are still running creative like it's 2023.
We tested 4 image models on the same ad — here's what won
Flux Pro, Imagen 3, Grok, and Nano Banana Pro on the same brief. The winner wasn't the one we expected, and the loser was a surprise too.
LLM-written ad copy: a 6-month performance comparison
We ran human-written ads against AI-written ads on the same campaigns for 6 months. The results weren't what either side of the debate predicted.