Your Marketing Used to Be Expensive. That Was the Point.

Here's a stat that should trouble anyone running AI-generated ads: AI ad creative on Meta gets 12% more clicks than human-made creative. It also delivers an 8% lower conversion rate on purchases above $100, and a 17% premium perception gap in favour of the human version.

More attention. Less trust. Same budget.

This is the central paradox of AI marketing in 2026. The tools have never been better at producing content. The content has never been worse at being believed.

And the explanation isn't about quality. AI-generated copy can be polished, grammatically flawless, and technically on-brief. The problem is deeper than craftsmanship. It's about what marketing economists call costly signaling, and it explains why making your marketing cheaper might be the most expensive decision you make this year.

A Flower Is a Weed With an Advertising Budget

Rory Sutherland, Vice Chairman of Ogilvy UK, has spent thirty years arguing that human behaviour runs on psychology, not logic. One of his most provocative insights involves flowers.

"A flower is simply a weed with an advertising budget," Sutherland writes in Alchemy. Flowers pour energy into producing colourful petals and sweet nectar. This is biologically expensive. And that expense is precisely the point. Bees visit flowers because the investment signals that the nectar is worth the trip. A plant that could produce the same signal for free would be ignored, because a costless signal isn't a signal at all.

The same mechanism operates in business. "The meaning and significance we attach to something is felt in direct proportion to the expense with which it is communicated," Sutherland argues. A FedEx letter is assumed to be important because who would pay eight dollars to send something trivial? A wedding invitation on embossed card signals different importance from an email, even when the information is identical. Red Bull costs more, tastes worse, and comes in smaller cans than Coca-Cola. Traditional market research would have killed it. But the unpleasant taste and premium price are features, not bugs. They share the properties of an effective placebo: expensive, strange, and available in restricted doses.

This is costly signaling theory: the idea that the cost of a signal is itself the message. When a business invests visibly in its marketing, buyers infer confidence, stability, and quality. Not because the marketing is necessarily better, but because a business willing to invest significantly is making what Sutherland calls a "sunk-cost demonstration of faith in its own product."

Now consider what happens when AI reduces the cost of producing marketing to near zero.

The Signal Collapses

The numbers are staggering. 87% of marketers now use generative AI in at least one recurring workflow. 74% of newly created web pages in 2025 contained some AI-generated content. In just three years, AI has produced more content than humans managed to create and digitise in 10,000 years.

Sam Tomlinson, one of the sharpest marketing strategists writing today, identified the core problem in his newsletter: "AI has reduced the cost to create middle-of-the-bell-curve content to near-zero." He calls the result the explosion of mid. "When every brand sounds 'professional' and 'optimised,' the risk becomes drowning in a sea of sameness."

His framework for understanding this is elegant: surplus value decreases with adoption. When a marketing tactic is new and rare, it works disproportionately well. When 87% of your competitors adopt the same approach, the incremental gain approaches zero. AI content in 2024 was a competitive advantage. AI content in 2026 is the new minimum.

Sutherland's costly signaling theory predicts exactly what happens next. When the cost of producing a signal approaches zero, the signal stops working. A blog post that took thirty seconds to generate signals something about how much the business believes in the argument it's making. Not because the reader consciously thinks "this was AI-generated," but because Daniel Kahneman's System 1 processing is extraordinarily good at detecting pattern, effort, and authenticity below conscious awareness.

We've written before about how AI homogenises marketing. But the signal collapse goes further than looking the same. It's about buyers feeling something is off, even when they can't articulate why.

Your Buyers Are Already Reacting

The data on consumer trust in AI-generated content is moving fast, and not in the direction marketers want.

And buyers are getting better at spotting it. Animoto's January 2026 study found 83% of consumers can identify AI-generated video content. Of those who spot it, 36% say it lowers their trust in the brand.

But the most revealing finding comes from DigitalApplied's 2026 AI Ad Creative Benchmarks, which analysed AI versus human creative performance across Meta campaigns:

We've covered why clicks don't always convert to leads before. This is the AI-specific version of that problem. The click looks identical in your reports. The buyer's trust level behind it is fundamentally different.

Why "Better AI" Won't Fix This

The instinctive response is to make the AI-generated content better. More polished. More personalised. More indistinguishable from human work.

This misunderstands the problem entirely. Sutherland's costly signaling theory doesn't say expensive things communicate better. It says the expense IS the communication. A hand-written thank-you note doesn't convey more information than a text message. It conveys more investment. The reader processes that investment as care, seriousness, and commitment.

When you make AI content indistinguishable from human content, you haven't solved the signaling problem. You've hidden it. And Tomlinson predicts this hiding gets harder: "When every brand is pushing ads that look like a piece of influencer content, it all becomes noise. Audiences catch on. It stops feeling natural, native and authentic, and it starts feeling like a template."

Kahneman's work explains the mechanism. System 1 doesn't analyse marketing piece by piece. It pattern-matches. It compares what it's seeing against everything it's seen before. And when System 1 has been exposed to thousands of AI-generated ads, blog posts, and emails, it develops a heuristic: this feels like everything else I've been ignoring. Skip.

You can't solve a perception problem by improving the thing people have already learned to perceive negatively.

What Counts as a Costly Signal in 2026?

Byron Sharp's research at the Ehrenberg-Bass Institute shows that brands grow by being easy to think of and easy to buy. But "easy to think of" requires distinctive brand assets that stand out from the category norm. When every competitor's marketing is generated by the same models trained on the same data, distinctiveness becomes the scarcest resource in marketing.

The question for any SME: where can you invest in signals that remain genuinely costly to produce?

What This Means for Your Business

This isn't an argument against using AI. Use it for research, first drafts, data analysis, scheduling, and the hundred operational tasks where efficiency matters and signaling doesn't.

But when your marketing faces the buyer, ask one question: "Could my competitor produce this with ChatGPT in five minutes?"

If the answer is yes, it's not a signal. It's noise. And your buyers are learning to tune out noise faster than your AI is learning to produce it.

The businesses winning right now aren't the ones producing the most content. They're the ones producing content that carries visible evidence of investment. Specific numbers from real campaigns. Original data nobody else has. Non-obvious expertise that took years to develop. Creative that looks like someone actually cared.

Marketing used to be expensive. That wasn't a bug. It was the mechanism that made it work. AI hasn't removed the need for that investment. It's made the investment more valuable than ever, precisely because so few businesses are still willing to make it.

Further Reading

• Rory Sutherland on Costly Signaling in Marketing - how signal cost creates perceived value in advertising
• Klaviyo 2026 AI Consumer Trust Report - year-over-year data on declining consumer trust in AI content
• AI Ad Creative Benchmarks 2026 - CTR vs conversion data comparing AI and human creative on Meta
• Animoto: 83% of Consumers Can Spot AI Videos - consumer AI detection research and brand trust impact
• State of AI in Marketing 2026 - adoption rates and benchmarks across the marketing industry