CMOs are spending millions on AI ad automation. Algorithms that bid smarter, optimize faster, test creative variants. The problem: they're optimizing the wrong thing.
A new study from VDO.AI just dropped the data everyone pretended didn't exist. Ad format and creative relevance matter more than AI automation in driving engagement. Not close. Massively more. And the industry response has been silence because it breaks the narrative everyone sold.
Here's what happened. The team tested thousands of ad campaigns across Meta, Google, and TikTok. They measured performance on engagement, CTR, and conversion lift. The result: creative relevance and format choice accounted for 70 percent of performance variance. AI optimization accounted for 12 percent. Platform reached another 10 percent. The rest was noise.
This should have broken ad tech. It didn't. Instead, the ad tech industry doubled down on automation because that's what investors fund. And CMOs kept buying it because the pitch deck looks smarter than "hire better creatives."
The Relevance Trap
Here's where this gets dark. Most CMOs already know their creative team matters. They can feel it. But they can't measure it in a spreadsheet. They can't fire it. And they can't outsource it to a vendor.
So they outsource what they can measure: the budget allocation, the bidding, the variant testing. Automation tools promise to handle the heavy lifting. And they do. But they're automating away the one thing that actually matters.
It's like hiring an accountant to do your sales job because you can measure what they invoice. The math looks good. The revenue goes down.
The VDO research breaks it down by format:
Video ads (especially on social) drove 3.2x higher engagement when creative matched the platform's native format over non-native formats. AI couldn't overcome format mismatch.
Shopping ads on Google and Meta performed 2.1x better when the product image and copy reflected what the actual searcher intended to buy. Relevance, not optimization, drove the lift.
Carousel ads on Meta underperformed by 40 percent when AI testing rotated through irrelevant variants. Humans picking the right order beat algorithm rotation.
The common thread: the moment the creative wasn't relevant to the intent, AI couldn't fix it by optimizing harder.
Why CMOs Keep Missing This
CMOs inherited a tech stack built on measurement. Pixels, events, conversion tracking. That tooling can measure if an ad appeared and if someone bought. It cannot measure if the creative was relevant.
So the incentive structure is backward. You can measure if your AI tool found a cheaper bid. You cannot measure if your human creative matched the moment.
This creates a perverse outcome. CMOs invest in tools that deliver measurable cost savings. Those tools commoditize the creative by testing thousands of variants. The commodity creative underperforms. So they buy more targeting and more AI to compensate. Budget goes up. Efficiency goes down. The AI gets blamed. But the system was rigged.
HubSpot's new Breeze AI is a perfect example. It auto-generates, tests, and iterates creative variants at scale. It works. It also guarantees that your best-performing ad is better than 99 percent of your worst performing ads, and worse than 1 percent of your best ads. You're optimizing the bottom. You're not reaching the top.
The Format Angle Most CMOs Ignore
Platforms are shipping new ad formats every quarter. Performance+, Advantage+, Direct Offers, Business Assistant, Performance TV. Each one promises better results if you turn on automation.
What the platforms don't say: the format itself is doing the heavy lifting. You'd get 60 percent of the benefit by just putting human-created, platform-native creative into the right format. The AI gets you the other 40 percent.
But formats require thinking. They require the creative team to understand where the ad appears, who sees it, and what they're trying to do in that moment. That's hard. So CMOs skip it and let AI pick.
Result: an ad that could run at 60 percent efficiency looks like it's running at 40 percent, because you measured the wrong thing.
Meta reported this quietly in their Q1 2026 earnings. Teams that built native vertical video (9:16) first, then optimized placement, saw 2.7x better CPL than teams that started with creative and let Meta pick the format. The format choice mattered more than the optimization.
What to Actually Do
If you're a CMO and you're not sure what to fix first:
Stop auto-generating creative at scale. You're not going to out-optimize the relevance problem. Pick formats (TikTok vertical, YouTube Shorts vertical, Meta feed square, etc.). Build native creative for each. Use AI to optimize variants of the native versions, not to auto-generate from a single asset.
Measure creative relevance, not creative quantity. Set a baseline. For one format, run a human-created ad and an AI-generated variant. Track performance per thousand impressions. You'll see immediately what your creative team is worth.
Own the format choice. Platforms want you on their newest format because they own the inventory. Your job is to see which format your audience actually lives in. That might not be Advantage+. Build for where they are. Then optimize.
Separate creative work from optimization work. Your creative team should never see the bidding strategy. Your optimizer should never rewrite creative without human approval. The system breaks when you let one person do both badly.
The Real Story
The reason this study landed without incident is that it threatens three parties equally: ad tech vendors (their automation is less valuable), platforms (their auto-format routing is less valuable), and CMOs (their vendor selection was wrong).
Easier to ignore it. Vendors keep selling automation. Platforms keep launching formats. CMOs keep wondering why their efficiency is stuck.
The out is uncomfortable. Better creative, not more AI. Better format thinking, not more testing. Better relevance measurement, not better cost measurement.
That's not a vendor you can buy. That's work you have to do.