AI-Made Ads Are Beating Human-Made Ones

A few days ago, a friend who runs paid media asked me out for tea. He looked miserable. His boss, he said, wants creative output to grow tenfold this year — and his team is just a handful of people. What was he supposed to do?

I asked him: you're still having humans make creative assets one by one?

He bristled. AI-made ads? Who'd fall for those? Users see through them instantly, all you get is curious clicks, and conversions collapse.

I've heard that line too many times. In 2025, Columbia University and Realize ran a study analyzing more than 500 million ad impressions. The conclusion stings:

AI-generated ads got a click-through rate of about 0.76%. Human-made ads, about 0.65%.

AI won.

And it didn't win by looking novel. The researchers specifically checked downstream conversions and found that the traffic AI ads brought in wasn't low quality. In other words, the worry that "AI ads can only bait curious clicks" was flatly demolished by the data.

Even more interesting was the secret sauce. The study found that AI ads users judged as "made by humans" performed best. AI hadn't learned to show off — it had learned to nail the honest creative fundamentals, like human faces in close-up and clean composition, more reliably than people do.

My friend stopped drinking his tea.

Not Using AI Isn't "Conservative" Anymore — It's "Falling Behind"

What does falling behind look like?

McKinsey's numbers: in 2024, 78% of companies were already using AI in at least one business function; by 2025, that figure had climbed to 88%.

Which means marketing teams still debating "should we use AI?" have only 12% of the room left. You're not making a forward-looking decision — you're taking a makeup exam for a course everyone else already passed.

A few more numbers:

Over 80% of marketing teams use generative AI; 93% of CMOs say they've seen clear ROI from GenAI; companies using AI see marketing ROI 20% to 30% higher than traditional approaches.

In marketing, AI is no longer a bonus — it's the passing grade.

The tools are already lined up on the shelf. Adobe Sensei and Google's ad AI adjust bids in real time; Jasper, Copy.ai, and Midjourney cover everything from ideas to design; HubSpot and ActiveCampaign have brought AI into lead nurturing and segmentation. Performance ad platforms like Realize go further, handling the whole pipeline — audience targeting, landing pages, real-time campaign optimization.

But let me offer one caution. Buying the tools is easy; the hard part is how the people on your team think about them. The moment many bosses mention AI, employees get nervous: is this here to replace me?

You need to make one thing clear: what AI takes over is repetitive work like publishing posts and pulling reports, and the time it frees up is for people to do judgment, creativity, and the things AI can't do. People use AI; they're not used by AI. If that doesn't click, the tools you buy will just gather dust.

Personalization Went From "Nice-to-Have" to "Basic Courtesy"

Have you ever had this experience: you search for running shoes on one app, open another app, and it's a screen full of running-shoe ads. Annoying.

Now a different scenario: you're browsing a shopping site late at night, and instead of pushing bestsellers, it shows you the pair you looked at three times last week and hesitated on — with a note: "40 yuan cheaper than when you last viewed it."

You bought it.

Both are personalization. What's the difference? The first is harassment; the second feels like it gets you. And consumers have voted with their attitudes: 71% of them expect personalized experiences from brands, and 76% express dissatisfaction outright when brands fail to deliver.

Brands are piling in too. 79% of marketers use AI for personalized content and targeting; customized call-to-action buttons can convert about 202% better than generic ones; 82% of consumers will even trade personal data for a better experience.

But here's a particularly ironic number: 85% of companies believe they deliver personalization, while only 60% of consumers agree.

What's the missing 25%? It's the gap between "I think I know you" and "you actually know me." AI can guess what you want to see from browsing behavior, device, location, and time of day — Netflix's and Amazon's recommendation engines have taken this to the extreme. But the more accurate the guesses, the more uneasy users get: how do you know I wanted this? What are you doing with my data?

So the deeper a brand goes on personalization, the more it must put three things out in the open: a clean opt-in mechanism, privacy settings people can actually understand, and an AI-usage explanation that doesn't play mind games. The EU's GDPR and California's CCPA draw the same line.

Personalization is an amplifier; trust is the principal. Lose the principal, and the amplifier just amplifies noise.

Automation: Hauling People Out of the Grunt Work

Here's a plain fact: less than half of a marketer's day may actually be spent "doing marketing." What's the other half? Pulling data, sending emails, adjusting budgets, watching dashboards.

AI can take over all of it.

79% of marketers already automate customer journeys, and half use it daily; among those who've adopted automation, 80% have seen lead growth. Email was the first battleground to fall: 71% of marketers use AI for email automation — subject lines generated automatically, send times matched automatically, an automatic follow-up whenever a user takes action.

Ad buying even more so. Optimization pros used to set alarms to get up at midnight and adjust bids. Now algorithms change bids, shift budgets, and swap placements in real time — faster than you, and calmer than you. Customer service too: chatbots field entry-level questions, and colleagues only handle the genuinely tricky ones.

Sounds great. But don't rush to push people out the door.

However beautiful AI's reports are, someone still has to glance at "is this anomaly real." However handy AI's headlines are, someone still has to gatekeep the brand voice. On content, AI's consistency is a known weakness: today it writes like you, next week it may drift off — and it will also confidently spout nonsense.

Every piece of content you publish must pass a human pair of eyes first. That's not distrust of AI — that's being responsible for your own brand.

The Real Divide Is in the Places You Can't See

Everything above is about what AI can do for you. But in 2026's marketing industry, the real dividing line isn't efficiency — it's responsibility.

AI was fed decades of humanity's books, web pages, and photos. Whatever went in comes back out. Bias, privacy, copyright, deepfakes, water and power consumption — every one of these is a real problem. Generative AI "hallucinates," fabricating facts with full confidence; AI-synthesized photos and videos have already been used in scams.

Regulation is catching up too. The EU's AI Act has entered phased enforcement, with disclosure, transparency, and risk-tiering requirements getting ever more detailed. In the US, there's no unified federal framework yet, but state-level transparency and consumer-protection bills keep coming, one after another.

Companies are doing their homework too, though unevenly: 75% of organizations have some kind of AI usage policy, but only 36% have built formal governance frameworks; three in five companies monitor their AI systems for fairness and bias; among US-listed companies, 36% already single out AI as a risk factor in their SEC filings.

What's a risk factor? The kind written into a prospectus, the kind you have to answer for to every shareholder.

Efficiency decides how fast you run this quarter; governance decides whether you can keep running.

Same for individuals. Once AI takes over pulling reports and doing segmentation, where does a marketer's value move? Toward oversight, toward judgment, toward being the ethics gatekeeper. New crafts like prompt engineering and data literacy — learn them early, claim your spot early.

Back to That Cup of Tea

After I told my friend all this, he was quiet for a moment, then asked: so where do I start?

I said, forget "tenfold" for now. Pick one step and hand it to AI as a trial — say, first drafts of creative assets, with humans doing only the final review. Then judge the results with the mindset of that Columbia study: don't guess, look at the data.

Can AI-made ads beat human-made ones? The data has already answered that. And whether you can use it fast and steady — that answer isn't in AI's hands. It's in yours.

Here's to him and his tenfold target this year — may it go smoothly.