Do AI ads actually work?

Short answer: yes, but not the way most people think.

The longer answer requires separating two very different things that get lumped together under "AI ads." There's AI-generated creative, which is what tools like Flayr, Arcads, and Creatify produce. And there's AI-optimized targeting, which is what Meta and TikTok's algorithms do on the platform side. Most people asking "do AI ads work" are really asking about the first one: can an ad made by AI actually get people to buy stuff?

What "working" actually means

An ad works when it generates more revenue than it costs to run. In paid advertising terms, that means a positive ROAS (return on ad spend). If you spend $100 on an ad and it generates $300 in sales, that's a 3x ROAS and the ad works. The creative is one variable in that equation. Targeting, offer, landing page, product-market fit, and pricing all matter too. But creative is the one variable you can change fastest, and it's the one that determines whether someone stops scrolling or keeps going.

So the real question isn't "do AI ads work" in some abstract sense. It's "can AI-generated creative stop the scroll and get a click as effectively as human-made creative?"

The honest answer by format

Photo ads are where AI creative is strongest right now. Tools using GPT Image produce product photos and lifestyle shots that are genuinely hard to distinguish from professional photography. We've shown Flayr-generated photo ads to media buyers without telling them the creative was AI, and most couldn't tell. For static image ads on Meta, AI creative performs at or near parity with human-made creative in most DTC categories.

Short-form video (under 15 seconds) is close behind. AI UGC tools using current generation video models like Seedance 2.0 produce talking-head clips and reaction videos that pass in a fast-scrolling feed. The quality isn't identical to a real creator, but for direct response advertising where the hook matters more than production polish, the performance gap is small.

Long-form video (30+ seconds) is still weaker. The longer the video, the more chances for something to look off: a weird hand movement, lip sync drift, an expression that doesn't match the voiceover tone. For ads that need sustained attention over 30-60 seconds, real creators still have an edge.

The volume advantage matters more than any single ad

Here's the thing that changes the math entirely. Even if an individual AI ad performs 20% worse than a human-made ad on average, the volume advantage more than compensates.

A freelance UGC creator costs $150-$500 per deliverable. At that cost, most DTC brands can afford to test maybe 3-5 creatives per month. An AI tool like Flayr costs $49.99/month and can generate 20+ creatives in the same period, plus delivers automated fresh creative every week.

The brand testing 20 variations finds their best performer faster, learns which hooks work in their niche sooner, and replaces dying creative before fatigue sets in. The brand testing 3 variations is always behind, always guessing, always running stale creative longer than they should.

Performance marketing isn't about making one perfect ad. It's about testing enough variations to find the 2-3 that outperform everything else. AI makes that testing volume accessible to brands that could never afford it before.

When AI ads don't work

Being honest about this: AI ads underperform in a few specific contexts.

When the product requires a physical demonstration. If your ad needs someone unboxing, applying, or physically using your product in a way that proves it works, AI can't convincingly replicate that yet. Makeup tutorials, fitness equipment demos, food preparation: real humans still win.

When the brand is built on personal authenticity. If your entire brand story is about a real founder, a real family recipe, a real community, then AI creative feels incongruent. The audience can sense the disconnect between "our brand is deeply personal" and "our ads are made by robots."

When the audience is highly sophisticated about AI. Some communities (tech, design, certain B2B segments) have trained their eye to spot AI content. For these audiences, being caught using AI creative without disclosure can damage trust.

The smart approach

The brands getting the best results with AI ads aren't using AI for everything. They're using AI for volume and testing, then doubling down on human creative for proven winners.

The workflow looks like this: generate 15-20 AI ad variations. Test them all with small budgets. Identify the 2-3 hooks and structures that outperform. Then hire a real creator to produce the premium version of those winning concepts. You get the testing efficiency of AI and the authenticity of human creative for the ads that actually scale.

This is essentially how Flayr's competitor intelligence system is designed to work. It analyzes which ad formulas are already proven to work for competitors, generates creative based on those formulas, and the weekly drop system ensures you always have fresh variations to test. When something wins, you know exactly which hook and structure to hand to a real creator for the hero version.

Do AI ads work? Yes. But they work best as a testing engine that finds winners, not as a replacement for every piece of creative your brand will ever need.