AI in Advertising: How to Use It Without Blowing It

I saw a number recently that stopped me cold.

StackAdapt and Ascend2 ran a study and found that only 39% of advertising agencies have actually integrated AI into their daily workflows. Another 18% have barely touched it.

My first reaction: That can't be right. The whole world is saying AI is revolutionizing advertising — how is actual adoption this low?

Then I thought about it, and it made sense.

The problem isn't the tools. There's no shortage of those. The problem is that most teams don't know where AI should sit in their workflow — and once it's there, where humans should stand.

Let's talk about that.

What Can AI Actually Do in Advertising?

Let's get the basics straight first.

AI's role in advertising breaks down into a few buckets:

Helping you see your audience. You used to box people in by age, geography, income. Now privacy regulations keep tightening, cookies are on their way out, and that road is narrowing. AI can analyze page content, semantics, and sentiment in real time, so your ad shows up next to what the user is reading right now. Ned Dimitrov from StackAdapt said something I think nails it: "The ad appears not because you know the user, but because the content of this page is right."

Helping you do the math. Predictive models can crunch signals at massive scale to estimate: How will this campaign likely perform? What's the right bid? What's the probability of conversion? You don't have to wait until it's over to do a post-mortem — you have a solid read before you even spend a dollar.

Helping you do the work. Creative version management, bid adjustments, budget allocation, report analysis — these tedious, time-consuming tasks are exactly what AI takes off your plate, freeing people up for real work.

Helping you watch for safety. Fake traffic and brand-unsafe content are perennial headaches in programmatic advertising. AI can do semantic analysis at the page level and catch risks before an ad is ever served.

Helping you make creative. This is the busiest battlefield of 2026.

Creative Production: Where AI Pays Off Immediately

Let me tell you something.

James Targett, Creative Project Manager at StackAdapt, shared a figure: in the time it used to take to write a single ad headline, they can now produce 50.

But here's what's even more striking: they tracked the click-through data, and the AI-written copy regularly outperformed both the client-provided versions and the internally hand-written ones.

That's not what surprised me most, though.

What surprised me most was a joint study. Columbia, Harvard, TU Munich, and Carnegie Mellon teamed up and found that AI-generated ads had a CTR of 0.76%, compared to 0.65% for human-written ones.

AI won.

But pay attention, because here's the key turning point: ads that look like AI generated them — whether or not they actually did — consistently performed worse.

In other words, what sinks you isn't AI itself. It's the "AI feeling."

Consumers can tell.

Targett put it in a way I fully agree with: The right way to use AI is to generate volume fast, then use human hands to curate, to refine, to set the tone. Collaboration, not replacement.

AI handles volume; humans handle taste.

Dynamic Creative Optimization: One Asset Becomes a Thousand

Here's another play I think is genuinely clever.

Dynamic creative optimization — the industry calls it DCO.

What is DCO?

You have an ad creative. The headline, image, call to action, product recommendation — these are all components that can be pulled apart. What DCO does is reassemble those components on the fly based on each viewer's real-time intent and context.

One creative, instantly spinning into hundreds or thousands of variations.

StackAdapt's platform data tells the story well: ads using DCO saw CTR up 32% and cost per click down 56%.

There's an even more interesting case. Vallo Media used DCO to dynamically adjust product ads based on users' browsing behavior, specifically targeting people who'd abandoned their shopping carts. The result? CTR jumped 60%, and this slice of advertising — just 12% of total budget — drove 30% of attributed ad revenue.

12% of the budget. 30% of the revenue.

The math speaks for itself.

Consumers Are Changing Their Minds

But it's not that simple.

In 2023, nearly 60% of consumers said they were comfortable with brands using AI in advertising. By 2024, that number dropped to 46%. Later that same year, nearly two-thirds of consumers felt uneasy about AI's use in ads.

What's worth paying attention to: this wariness is creeping into younger demographics. A 2025 report from a research firm found that 39% of Gen Z dislikes AI-generated ad creative — nearly double the rate among millennials.

Young people are supposed to be the ones who embrace new tech fastest. When even they start pushing back, you need to take it seriously.

Why?

Because what AI makes sometimes falls into the "uncanny valley" — it looks human-made, but something's slightly off. And that's more unsettling than something obviously machine-made.

A global survey of decision-makers found that 54% of marketing executives worry that over-reliance on AI will erode the kind of creativity that genuinely moves people in advertising.

Te'Shawn Dwyer from StackAdapt's creative studio said it well: AI is incredibly useful at the ideation stage — quickly generating concepts, sketching storyboards. But when brands skip the human curation step, the result feels cold and lifeless.

AI handles quantity; humans handle quality.

So, How Do You Actually Implement This?

Alright, enough theory. You might be asking: what do I do when I go to work tomorrow?

Matt Travers, Managing Director of BRAIVE, an AI consulting firm in Australia, offered a great starting point:

"Ask yourself one question first: For me, where exactly does it hurt the most?"

Start from the pain point. Not from the tool.

Then break that pain point apart — where does the process break, where does it bottleneck — and use AI to unblock the most critical link.

For most teams, that starting point is creative production, targeting, bidding, or performance measurement. It's wherever you're slowest and most error-prone when doing it manually.

How do you actually move forward?

Step one: Define the goal. AI works best when it solves a specific problem. Is your targeting not precise enough? Is creative output too slow? Figure out what you actually want, so you can judge whether AI is helping.

Step two: Choose the tool. Don't ask which platform has the most features. Ask which one fits into your existing workflow. All the features in the world mean nothing if you can't actually use them.

Step three: Test before you scale. Take one variable, one goal, one clearly defined evaluation window, and run a controlled test. Don't roll it out across the board on day one. Product demos and real-world performance data are frequently two very different things.

Step four: Train the team. LinkedIn's B2B marketing benchmark report found that 43% of marketers cite "lack of internal AI skills" as the biggest barrier to adopting generative AI. It's not that the tools aren't good enough — it's that people don't know how to use them. How to write prompts, how to evaluate outputs, when to bring in human judgment — all of these require practice.

Step five: Don't get greedy. The most common mistake is trying to do too much at once. A scattershot "use AI for everything" approach only leads to fragmented strategy and muddy results. Run small, document, iterate.

The Pitfalls You Can't Avoid

The more you use AI, the higher the odds of something going wrong.

In the StackAdapt and Ascend2 study, the top concerns agencies flagged: data privacy and compliance (32%), brand safety and ad fraud (27%), the balance between automation and personalization (27%), and ethical issues (18%).

IAPP ran a broader survey — spanning the Association of National Advertisers to Salesforce to IAB — and the risks that kept surfacing included: algorithmic bias, hallucinations, data privacy, uncertainty around AI-generated content, and intellectual property.

These are all real problems.

What's the fix?

Keep humans in the loop.

Yang Han, Co-founder and CTO of StackAdapt, said something I find remarkably precise: "The core question isn't AI versus humans — who replaces whom. It's how the two work together. AI is great at data-driven tasks, automation, predictive analytics. But it lacks human intuition, creativity, and ethical judgment."

Salesforce, PwC, and IAB are all emphasizing the same thing: set boundaries on what AI is allowed to generate, audit outputs regularly after deployment, and train your team to verify results and escalate risks.

But here's a deeply unsettling data point. IAB's report found that over 70% of marketers have encountered AI-related issues — hallucinations, bias, off-brand content — yet fewer than 35% plan to increase investment in AI governance or brand integrity in 2026.

Things are going wrong. But the defenses aren't being reinforced.

That might be the most dangerous mindset in the industry right now.

The Real ROI

Having talked about all these risks, how does AI's math actually work out?

StackAdapt's "State of Programmatic Advertising 2026" report includes a set of internal platform data:

Ads using first-party data or AI-powered contextual targeting achieved ROAS up to 2x higher than third-party targeting. Ads using DCO saw CTR up 32% and CPC down 56%.

McKinsey's research found that 24% of marketing and sales teams achieved revenue growth of 6% or more in the past year thanks to AI. StackAdapt's personalization study showed that 93% of brands and 94% of agencies believe AI is improving the speed and efficiency of programmatic marketing.

Here's another number worth sitting with.

StackAdapt's research found that among brands that fully integrated AI across channels, 79% could more accurately measure the revenue impact of personalization. Among brands that don't use AI at all? That number is 14%.

79% versus 14%.

That's not a rounding error. That's a different league.

Looking Ahead: The Trends to Watch

Let's close with trends. And to be clear — these are things already happening, not pie-in-the-sky.

First, LLM advertising is moving from the margins to the mainstream. People are increasingly accustomed to searching, comparing, and making decisions inside tools like ChatGPT. McKinsey says half of consumers already use AI-driven search to research and discover products. EMARKETER predicts that in 2026, AI chatbots will assist 63.3 million US consumers. Global LLM advertising spend is projected to exceed $101 billion by 2030. This is happening now.

Second, agentic AI is taking over more of the execution layer. Media buying is shifting from "manual operations" to "strategic orchestration." IAB's data shows that two-thirds of US ad buyers plan to pay more attention to agentic ad purchasing this year. Work that used to take weeks — data analysis, media planning, reporting — agentic AI can handle in minutes.

Third, brand safety is shifting from "risk avoidance" to "unlocking new inventory." Programmatic podcast advertising is a textbook example. Without transcribed text, it used to be impossible to evaluate at scale whether a show's content was suitable for an ad placement. Now AI can analyze the themes, context, and risk of hundreds or thousands of episodes simultaneously. Inventory that was previously untouchable becomes investable once it's been assessed.

Fourth, orchestration is becoming more predictive. AI-generated bidding strategies, budget reallocation recommendations, and performance simulations are increasingly common. Humans set goals and constraints; AI runs the simulations, spots opportunities, and guides decisions. The shift is from "reacting after the fact" to "anticipating before the fact."

Fifth, retail media measurement is starting to mature. Global retail media ad spend is projected to grow 52.6% over the next five years, but measurement standards have been fragmented and opaque. AI is making "why did this ad perform this way?" answerable — not just "did this ad perform?" 86% of commerce media decision-makers say improving measurement and attribution is a high priority for next year.

A Final Thought

Vitaly Pecherskiy, Co-founder and CEO of StackAdapt, once made an observation that I think is the key to understanding all of this change.

Over the past decade-plus, the marketing technology space has produced a proliferation of "leaders" — marketing automation, CDPs, programmatic advertising. But they don't talk to each other. They're all silos.

In 2011, Scott Brinker's marketing technology landscape had 150 products. The 2025 edition: 15,384.

AI can do something within each of these silos. But the real value emerges when these systems start connecting — when AI can make decisions across the boundaries of planning, creative, media buying, and measurement.

What AI can do when it connects everything matters far more than what any single AI can do.

That, I think, is the direction most worth watching.

But here's the thing — no matter how far AI goes, one thing probably won't change.

LinkedIn's report says AI skills and human skills grow together. People who learn AI are simultaneously practicing adaptability to change, trust-building, and logical reasoning.

Advertising is probably the same. People who know how to use AI will very likely replace the ones who don't.

May you be the former.