What If AI Makes Your Marketing Cheaper — and Your Business Worse?

Generative AI is very good at making things: faster, cheaper, and at scale. For marketing leaders, that sounds like an obvious win. More concepts, more versions, shorter production cycles, and lower creative costs.

But a new study raises a more uncomfortable question:

What if AI makes your marketing cheaper — and your business worse?

In Research: AI-Generated Ads Perform Worse Than Human-Made Ones—Even When Customers Can’t Tell Them Apart, researchers working with Ipsos showed 3,000 U.S. consumers 20 video advertisements from major brands. Half were human-made; half were recreated entirely with AI from comparable creative briefs.

The AI ads did not look obviously bad, and consumers struggled to distinguish them from human-made versions. But the business-performance measures told a different story. On average, the AI-generated advertisements were 14% weaker on short-term sales potential and 17% weaker on long-term brand equity.

---

Good Enough to Approve

The danger is not necessarily terrible AI creative—that would be rejected before it ran. The difficult problem is work that looks perfectly acceptable: polished, on-brand, professional. Good enough to approve.

The study describes exactly this risk: AI can lower the production cost of creative that moves smoothly through internal review while still quietly sacrificing performance in the marketplace. The core question is no longer “Does this look like AI?” but “Does this work?”

Everyone Can Be Right — and the Business Can Still Lose

Imagine the campaign review. All three can be correct about the shift toward AI creative:

The CFO sees: Lower production cost.

The CMO sees: Faster campaign cycles.

The CEO sees: More AI adoption.

But none of those metrics reveal whether the advertising became more persuasive. The study warns that organizations can gradually reward volume, speed, and short-term savings while weakening the actual business effect. AI can improve the efficiency of producing an output while reducing the effectiveness of the outcome.

Are We Measuring the Output or the Outcome?

Suppose an AI campaign cuts production costs dramatically. If we measure easy production metrics, the AI system looks fantastic:

Cost per asset ↓ • Production time ↓ • Content volume ↑ • Number of campaigns ↑

But the researchers argue that these metrics can obscure the economics that actually matter. The economics look very different if we measure genuine business outcomes:

Revenue per asset • Conversion per media dollar • Persuasion • Consumer attention • Trust • Brand equity

This gives us a simple rule:

Measure the outcome, not merely the activity. The cheapest advertisement is not necessarily the cheapest business outcome.

But AI Didn't Lose Everywhere

This is where the research gets more useful: some AI-generated advertisements performed quite well. The lesson is not human good vs. AI bad. Generative AI is exceptionally good at recognizing and recombining patterns.

AI matched human counterparts on short-term sales potential in cases that were product-driven, direct, and built around familiar, bounded structures. If the assignment is to localize an existing idea, adapt a campaign for a new channel, or show a product demonstration, AI may be exactly the right tool.

Give AI the Work Where Sameness Is Efficient

The study found AI had more difficulty when the creative depended on emotional specificity, humor, cultural timing, narrative tension, or a surprising point of view—the factors that make advertising memorable.

This suggests a better division of labor. Humans strategize, insight, and build emotional connections. AI handles execution-heavy tasks like versioning, resizing, internal mockups, early storyboards, and first drafts.

Give AI the work where sameness is efficient. Give humans the work where difference creates value.

The Real AI Decision Is an Allocation Decision

The real question for management moves away from blanket adoption toward deliberate allocation. A structured Decision System determines how AI should be used based on context:

• If AI improves cost AND maintains equivalent business performance: Scale it.
• If AI cuts production costs but materially reduces conversion or brand value: Don’t use it.
• If AI is strong at execution but weak at creative concepts: Redesign the workflow. Let humans lead the strategy and use AI to multiply and execute it.

From AI Adoption to AI Economics

An AI system can be technically successful while being economically unsuccessful. It can generate exactly what it was asked to generate, reduce costs, and still fail to improve the outcome the business ultimately cares about.

That is why AI ROI cannot stop at the point where the model produces something—it must follow the output all the way to the business result. The real question is not “How much cheaper did AI make the work?”, but:

Did AI make the business better?