The best AI in advertising examples split into two groups. There are the stunts: flashy, viral campaigns like Heinz’s AI Ketchup or Coca-Cola’s AI Christmas ad. And there are the systems: quiet, behind-the-scenes uses like JPMorgan testing thousands of ad headlines or Netflix swapping thumbnails for every viewer. The stunts get the headlines. The systems get the results.
But they’re not the same thing. A brand generating a weird DALL-E image for social buzz and a brand using AI to test fifty ad variations? Completely different work. Understanding which type you’re looking at, and which one you can actually copy, is the whole game.
The examples that get headlines (AI advertising stunts)
Heinz AI Ketchup (2022). Heinz typed “ketchup” into DALL-E and found that every AI-generated image looked like a Heinz bottle. They turned it into a campaign: “even AI knows ketchup means Heinz.” It earned 1.15 billion views across platforms and a reported 2,500% return on investment. Clever concept, perfect brand fit.
Coca-Cola’s AI Christmas (2024-2025). Coca-Cola remade their classic Christmas ad with AI-generated visuals. The internet was not kind. Social media called it “soulless” and “creepy.” But independent ad testing firm System1 scored the 2025 version at 5.9 stars, their highest possible rating. What people say online and what people actually respond to are, apparently, two different things. Behind the scenes, Coca-Cola used over 70,000 generated clips and 5+ specialists. Not exactly cheap or fast.
Toys “R” Us Sora film (2024). The toy chain made a brand origin film using OpenAI’s Sora video generator. It got press for being a “first.” The actual video looked like a fever dream. Most of the coverage focused on how weird it looked, not how well it worked. (For more on where AI-powered marketing videos are headed, it’s improving fast.)
Kalshi NBA Finals (2025). A prediction market company spent about $2,000 to produce an AI-generated NBA Finals ad. Traditional production would have cost $50,000+. The ad ran on national TV. Quality was rough, but the cost savings were real. And Kalshi got millions in free press just for being the company that did it.
Svedka Super Bowl (2025). Svedka’s parent company ran an AI-generated Super Bowl spot. It scored in the bottom 3% for likeability according to iSpot ad testing. Sara Saunders, the CMO, later admitted the ad “misses the heart and soul.” When your own CMO calls your ad hollow, that tells you something.
My take: These stunts are a PR strategy, not an advertising strategy. Heinz worked because the AI output reinforced the brand. The rest mostly got attention for being weird. If your goal is “get people talking,” stunts can work. If your goal is “sell more stuff,” keep reading.
The examples that make money (AI advertising systems)
JPMorgan Chase + Persado. JPMorgan signed a five-year deal with Persado, an AI copywriting tool. The reason: Persado’s ad headlines got up to 450% higher click-through rates (the percentage of people who click on an ad after seeing it). Not 4.5% better. 450% better. The AI tested thousands of word and phrase combinations that no human team could run manually.
Netflix personalized thumbnails. Netflix uses AI to show every viewer a different thumbnail image for the same show, based on what that person has watched before. If you watch a lot of comedies, you might see the funny scene. If you watch dramas, you get the tense one. Netflix has estimated this personalization is worth over $1 billion per year in retained subscribers.
Unilever Beauty AI Studio. Unilever built an internal AI system that generates 400 product images per product instead of 20, at 87% lower cost. Their video completion rate (how many people watch to the end) doubled. This is AI doing the boring production work that used to take weeks of studio shoots.
Michaels (craft store) email personalization. Michaels used AI to go from 20% personalized email content to 95%. The result: 25% higher click-through rates on their emails. Not by writing better emails. By writing more versions and matching them to the right person.
Meta Advantage+ creative. Meta’s AI ad system automatically tests different combinations of your images, headlines, and descriptions, then pushes budget toward the combinations that perform best. Advertisers using it report 22% better return on ad spend (ROAS, the money you make per dollar spent on ads). You upload the raw materials, and the AI does the mix-and-match.
These examples share one thing: the audience never knows AI is involved. That’s the pattern. When AI is the quiet engine running the testing and personalization behind the scenes, it works. When AI is the headline of the campaign, results are a coin flip.
Why the stunts get attention but the systems get results
Ad executives and consumers see AI ads very differently. And the gap is getting wider. The IAB surveyed both groups in January 2026: 82% of executives think consumers feel positive about AI in ads. Only 45% of consumers actually do. That’s a 37-point gap, up from 32 points the year before.
It gets worse when you look at the brain research. NielsenIQ ran EEG brain scans (tests that measure electrical activity in the brain) on people watching AI-generated ads versus human-made ads. The AI ads created weaker memory encoding. In plain language: people literally remember them less. Even when the ads looked polished, the brain treated them as less important.
Kantar’s facial coding study found something similar but different. AI ads triggered stronger emotional reactions, but the emotions were more negative. People felt something. It just wasn’t good.
Think about it like this: a stunt is a firework. Bright, loud, everyone looks up. But nobody remembers which fireworks company made it. A system is a thermostat. Nobody notices it’s running, but the house stays the right temperature. The brands making money from AI advertising are building thermostats, not fireworks.
My take: The 37-point perception gap is the single most important number in AI advertising right now. Ad teams are wildly overestimating how much consumers like this stuff. That doesn’t mean AI in ads is bad. It means the visible AI is bad. The invisible AI works great.
How small teams use AI in advertising right now
Here’s what you can actually do today, without a million-dollar AI studio:
1. Creative testing at scale. Generate 20-50 ad image and copy variations using AI tools, run them on paid social, kill the losers, and put more money behind the winners. This is the Meta Advantage+ model, just done manually. You’re not trying to make one perfect ad. You’re trying to find out which version works by testing a lot of them cheaply. That’s how generative AI is used in advertising at the practical level.
2. Ad copy optimization. Use ChatGPT or Claude to write 10 different headlines for the same offer. Test them. The JPMorgan/Persado model is just this: generate variations, test, keep the winners. You don’t need a $10M Persado contract. You need a chat window and a spreadsheet.
3. Dynamic creative. Swap images and copy based on who’s seeing the ad. Cold audience gets a problem-aware headline. Warm audience gets a product-specific one. Most ad platforms now support this natively with AI-powered ad bidding and targeting.
4. Resize and reformat. You have one hero creative. AI can generate the 15 different sizes you need for Instagram Stories, LinkedIn feeds, display banners, and YouTube pre-rolls. The boring work that used to take a designer half a day.
5. Personalized messaging. Different copy for different customer segments. The Michaels model: the same product, pitched differently to different people. If you’re running email alongside your ads, AI for small business marketing covers how to set this up step by step.
If you want to see what AI marketing campaign generators can do for variation testing, they’re getting surprisingly good at the bulk production part. The strategy still has to come from you.
What separates AI ads that work from ones that backfire
After looking at dozens of real AI marketing examples, a clear pattern emerges:
AI ads work when:
- The content is functional (product shots, resizes, A/B test variations)
- The audience is warm (they already know you)
- The format is performance-based (paid social, display, retargeting)
- Nobody needs to feel emotionally connected to the ad
AI ads backfire when:
- The campaign depends on emotional storytelling
- The audience is cold (first impression)
- The brand makes AI the headline (“look, we used AI!”)
- Authenticity matters more than reach
Alaska Airlines found a nice middle ground: using AI-generated visuals for dream destinations. It worked because the premise was explicitly fantasy. Nobody expects a dream to look real.
The disclosure question is tricky. 73% of consumers say knowing an ad used AI doesn’t change whether they’d buy. But among Gen Z, 30% label AI-using brands “inauthentic” and 24% call them “unethical.” Academic research confirms that AI disclosure leads to less favorable attitudes, especially for premium brands.
Rachel Lyndon-Jones, CMO at Ouma, put it well: “cultural exhaustion from meaninglessness.” AI makes it easy to produce more content. But more of what doesn’t work is just expensive noise.
The decision rule is simple: use AI for production and testing. Keep human judgment for strategy and emotion. If you’re looking at the best AI tools for marketing or exploring AI tools for social media ads, focus on the ones that help you test and personalize, not the ones that try to replace your creative judgment.
Advertising is just one piece of the puzzle. The same stunts-vs-systems split plays out across email, content, and AI content strategy more broadly. If you want a wider view, there are 15 examples of AI across marketing worth looking at. And if you’re just starting out with AI in digital marketing, the advertising examples here show you what the stakes look like.
How I can help
You’ve just seen the difference between brands that use AI as a PR stunt and brands that use it as a growth system. The system side (testing more variations, personalizing at scale, cutting production costs) is where the real leverage sits. And you don’t need a Netflix-sized budget to start.
If you want help setting up AI-powered creative testing for your ads, that’s exactly what I do. I’ll help you figure out where AI fits your workflow, build the testing system, and let the data pick the winners instead of guessing.
FAQ
What are examples of AI in advertising?
AI in advertising examples fall into two categories. Stunts include Heinz’s AI Ketchup campaign (1.15B views), Coca-Cola’s AI Christmas ad, and Kalshi’s $2K AI-produced NBA Finals spot. Systems include JPMorgan’s Persado partnership (450% higher click-through rates), Netflix’s personalized thumbnails ($1B+ value), Unilever’s Beauty AI Studio (400 assets per product at 87% lower cost), and Michaels’ email personalization (25% more clicks on emails). The systems consistently outperform the stunts on business results.
How is AI used in ads?
AI is used in ads in two main ways. First, creating the ad: generating images, video, and copy variations from a brief instead of building each one by hand. Second, optimizing delivery: using AI for targeting, bidding, and personalization so the right ad reaches the right person. For the creative side, see how generative AI works in advertising. For the delivery side, see AI PPC management.
What are the best AI advertising campaigns?
It depends on what “best” means. The most talked-about AI campaigns are Heinz AI Ketchup and Coca-Cola’s AI Christmas ad. The most effective are JPMorgan + Persado (450% more clicks), Netflix thumbnails ($1B+ value), and Meta Advantage+ (22% better return on ad spend). The effective ones rarely make headlines because the AI is invisible to the viewer.
Do AI-generated ads actually work?
Mixed. AI-generated ads work well for testing and production at scale: generating variations, resizing creative, and personalizing copy. They underperform for emotional brand-building. NielsenIQ brain research shows AI ads create weaker memories, even when polished. Kantar facial coding found AI ads trigger stronger but more negative emotions. The format and use case decide the outcome, not AI itself.
Can small businesses use AI for advertising?
Yes, and they arguably benefit the most. Small teams can generate dozens of ad variations cheaply, test them on paid social, optimize copy with AI tools, and personalize messaging by audience segment. The Michaels model (95% personalized content, 25% more clicks) is doable at any scale. Check out AI for small business marketing and AI-enhanced content marketing for step-by-step guides.