A Professor’s Refrigerator Test
For three years, advertising professor Carrie Riby ran a simple experiment in her “Big Ideas in Advertising” class at Syracuse University. She asked her students to use AI to create an ad selling themselves. The results were technically impressive — polished, on-brand, often startlingly professional. But something odd kept happening: not a single student ever loved their AI-made creation enough to want it stuck on the family refrigerator, the classic marker of something you’re genuinely proud of.
That small, recurring observation turned into the seed of a much bigger question. If the students creating the AI ads were themselves underwhelmed by the output, why would anyone expect a random consumer to feel any differently when they saw it?
That question is now the subject of one of the most talked-about advertising studies of 2026 — and it has a direct, practical payoff for anyone anywhere in the world who is trying to build an income around AI skills, whether you’re a freelance copywriter in Bangkok, a solo marketer running Meta ads for a shop in Ho Chi Minh City, or a digital nomad pitching AI-powered content services from a co-working space in Chiang Mai.
Inside the Experiment: 10 Brands, 20 Ads, 3,000 Verdicts
The study was a collaboration between the global research firm Ipsos and two faculty members from Syracuse University’s Newhouse School of Public Communications, Adam Peruta and Carrie Riby, working alongside Ipsos creative excellence lead Ryan Barthelmes. The study paired existing human-made ads, produced before 2021 to ensure AI tools were not involved, with fully AI-generated counterparts built from the same strategic brief — the document ad professionals use to lay out a campaign’s objectives, messaging, and tactics.
In other words, this wasn’t a case of AI going up against a strawman. Both versions of each ad started from the exact same instructions. The only variable that changed was who — or what — executed them.
Ten brands were selected across sectors including consumer packaged goods, fashion, automotive, and technology: Cheerios, Chewy, Febreze, Fiat, H&M, Old Navy, Herbal Essences, Ray-Ban Meta, TurboTax, and Visa. That produced 20 total ads — one human-made, one AI-made, for each brand. Those ads were then shown to roughly 3,000 consumers in the United States, who were asked to rate them and, crucially, guess which version had been made by a machine.
If you make a living creating content, running campaigns, or advising clients on marketing strategy, this is essentially the most rigorous head-to-head test of your future competitor that has been run to date.
The Illusion Holds Up — Until You Check the Sales Data
Here’s the part that should give every marketer, freelancer, and small business owner pause: consumers genuinely could not tell the difference. Only 13% of viewers who saw an AI-generated ad felt at least somewhat confident it had been made by AI — and that figure was identical to the share of viewers who wrongly suspected a human-made ad was AI-generated. Roughly 40% of all viewers said they simply couldn’t tell either way.
Read that again. People were just as likely to mistakenly accuse a human-made ad of being AI as they were to correctly spot an actual AI ad. The visual and conceptual “uncanny valley” that used to separate machine-made content from human work has, for practical purposes, closed.
But looking convincing and performing well turned out to be two very different things. Using Ipsos’s sales-validated measurement system, the human-made ads outperformed the industry benchmark by 11 points on average, while the AI-made ads underperformed it by five points. That’s a real, measurable gap in predicted short-term sales impact — invisible to the eye, but very visible on a performance dashboard.
This is the finding that gives the study its title, borrowed almost verbatim from the Ipsos report itself: the real risk with generative AI advertising isn’t obviously bad creative — it’s work that looks good enough to get approved internally while quietly underperforming once it actually reaches the market. “Good enough” sailed through the approval meeting. It just didn’t sell as much product.
แนะนำ
Where AI Actually Wins: The Straightforward Sell
The study didn’t find that AI is universally weaker — it found that AI’s strengths and weaknesses are highly predictable. AI performed best when a brief was straightforward and product-driven, and struggled when the creative challenge required storytelling, emotional nuance, or a genuine point of view.
Think about what that means practically. A brief built around a clear, functional value proposition — the kind of ad you’d expect from a tax-preparation service walking through a refund benefit, or a payments network highlighting how fast and secure a transaction is — plays directly to what large language and generative models do well: synthesizing a message, matching a tone, and producing something clean, on-brief, and technically correct at high speed and low cost.
This is genuinely useful information if you’re trying to build a business or a freelance career around AI tools. It tells you exactly where AI-assisted work is currently strong enough to be sold with confidence: product descriptions, feature-benefit ad copy, direct-response variations for A/B testing, localized versions of an existing campaign, and high-volume creative for performance marketing where the measure of success is click-through rate rather than brand affinity.
Where Humans Still Win: The Cheerios Effect
The flip side of that finding is where the study gets interesting for anyone worried about being replaced. The single strongest result in the entire study came from the Cheerios pairing, where a deeply human-centered brief produced the highest combined effectiveness scores across both the human and AI versions. Cheerios, notably, is a brand whose advertising has long leaned on warmth, family, and emotional storytelling rather than product features.
That combination — a brief that demanded genuine emotional insight, married to human execution — produced the standout performance of the whole experiment. It’s a strong signal that the parts of advertising (and, by extension, content marketing, brand storytelling, and persuasive writing generally) that depend on empathy, cultural nuance, humor, and a distinctly human point of view are exactly the areas where paying a premium for a human creator still makes clear business sense.
Ryan Barthelmes, who led the project from the Ipsos side, put the industry-wide implication plainly: every chief marketing officer is currently asking whether AI can replace their creative agency, and this research offers a data-backed answer — AI is a powerful tool, but human storytelling and emotional connection still create a measurable competitive edge, and the future is humans and AI working together rather than one replacing the other.
What This Looks Like on the Ground in Southeast Asia
This research was conducted with American consumers and American brands, but the underlying dynamic — AI closing the gap on execution while the strategic and emotional layer stays stubbornly human — is playing out just as fast across Southeast Asia’s fast-growing digital economy, including in Thailand.
Consider the everyday tools of small-business marketing here. Many Thai retailers run their entire customer relationship through a LINE OA (LINE Official Account) — a business messaging profile on LINE, the dominant chat app in Thailand, used for everything from broadcasting promotions to handling customer service chats. Increasingly, those broadcast messages, product captions, and reply scripts are drafted with AI assistance rather than written from scratch. Similarly, sellers on Shopee and Lazada, the two largest e-commerce marketplaces across Southeast Asia, routinely use AI tools to generate product photography backgrounds, translate listings, and draft short-form ad copy for in-app promotions.
แนะนำ
And on Fastwork, one of Thailand’s largest local freelance marketplaces (its regional equivalent to Fiverr or Upwork), it’s now common to see gig listings explicitly built around AI-assisted services — AI product photography, AI-written product descriptions, AI-edited short video ads. Demand for these gigs is real precisely because they match the category the Syracuse-Ipsos study flagged as AI’s strength: fast, functional, benefit-driven content at a lower price point than a full creative team.
What’s largely absent from that same freelance marketplace, and where rates remain highest, is anything requiring a genuine strategic or emotional read on a brand — the campaign concept behind a Songkran promotion that actually connects with how Thai families experience the holiday, or a brand story for a hospitality business that needs to feel authentically local rather than machine-translated. That is the Cheerios lesson showing up in a completely different market: execution has been commoditized, but insight has not.
The Freelancer’s Playbook: Turning This Data Into Income
For anyone actively trying to earn a living around AI tools — whether that’s a full-time freelance career, a side hustle alongside a day job, or an agency built on AI-assisted workflows — this study is close to a strategy document in disguise.
Sell AI-assisted execution as a volume product, not a premium one. The data shows this work is genuinely good — reliably good enough to pass approval — but it’s priced accordingly by the market. Position it as fast, scalable, and cost-efficient: multiple ad variants for testing, localized versions of an existing campaign, high-frequency social captions, product listing copy at scale.
Sell strategy and emotional insight as the premium tier. The Cheerios result is the single strongest argument available right now for why a human creative brain — one that understands a specific audience’s culture, humor, and emotional triggers — is worth paying more for. If you can demonstrate that kind of insight, price it like the competitive advantage the data says it is.
Package yourself as the judgment layer, not just the prompt-writer. The Harvard Business Review framing of this research is telling: the real danger with AI creative isn’t bad output, it’s mediocre output that looks polished enough to sail through approval unchecked. That makes a skilled human reviewer — someone who can catch “good enough” before it goes live and knows when to push for something better — an increasingly valuable and sellable role in its own right, distinct from either pure creative work or pure AI operation.
Use the “indistinguishable but underperforming” finding as a sales pitch to clients. Small business owners and marketing managers are exactly as confused about this trade-off as the consumers in the study were about telling AI ads apart. Being able to explain, clearly and with data, when AI is the right tool and when it isn’t, is itself a marketable consulting skill — one that doesn’t require you to be the world’s best copywriter, just genuinely informed.
The Takeaway
The uncomfortable and genuinely useful truth in this research is that AI has already crossed the threshold of looking real. It has not yet crossed the threshold of working as well as a human at the hardest, most valuable part of advertising: making someone feel something specific enough to change their behavior. For anyone trying to make money with AI tools — in Thailand, across Southeast Asia, or anywhere else in the world — the opportunity isn’t in pretending AI can do everything. It’s in knowing precisely which half of the job it has already mastered, pricing that accordingly, and building your actual career around the half it hasn’t.
Key Takeaways
- A rigorous 3,000-consumer study found people could not reliably tell AI-made ads from human-made ones, even when directly asked to guess.
- Despite that, human-made ads outperformed a sales-validated benchmark by 11 points on average, while AI-made ads underperformed it by 5 points.
- AI performs best on straightforward, product-driven, benefit-focused briefs — the kind of work that scales easily and is already being commoditized on freelance platforms across Southeast Asia.
- Human creators still hold a clear, measurable edge whenever a brief demands storytelling, emotional nuance, or genuine cultural insight, as shown by the standout Cheerios result.
- The smartest way to build income around AI right now is to price execution as a fast, scalable commodity and price strategic or emotional insight as the premium, human-only tier.
Frequently Asked Questions
Q: Can AI actually make an ad as good as a professional creative team?
A: It can produce something visually and technically convincing enough that consumers usually can’t tell it apart from human work, but sales-validated testing shows it still underperforms human-made ads on average, especially for anything requiring emotional storytelling.
Q: What was the Ipsos and Syracuse University AI advertising study?
A: It was a 2026 study pairing 10 real brands’ human-made ads with AI-generated versions built from the same creative brief, then testing all 20 ads with roughly 3,000 U.S. consumers to compare recognition and predicted sales performance.
Q: Could consumers tell which ads were made by AI?
A: No — only 13% felt confident an AI-made ad was actually AI-made, which was the same rate at which people wrongly suspected a human-made ad of being AI-generated.
Q: If people can’t tell the difference, why did AI ads perform worse?
A: Sales-validated scoring showed AI ads underperformed the industry benchmark by five points while human ads over-performed it by eleven, meaning the gap shows up in predicted buying behavior rather than in how the ad looks.
Q: What kind of advertising is AI actually good at right now?
A: AI performs best on straightforward, product- and benefit-focused briefs, such as functional value propositions, rather than ads that need to build an emotional connection.
Q: What kind of advertising still needs a human?
A: Campaigns built around storytelling, empathy, humor, or a distinct emotional point of view still perform best with human creative input, as shown by the study’s top-performing brand pairing.
Q: Is this relevant to small businesses in Thailand, not just big U.S. brands?
A: Yes — the same split is already visible locally, with AI handling high-volume tasks like LINE OA broadcast messages and Shopee or Lazada product listings, while brand storytelling and culturally specific campaigns remain human-led.
Q: Can freelancers actually make money using AI for advertising work?
A: Yes, particularly by offering fast, scalable AI-assisted execution — such as ad variants, product copy, or localized content — as a distinct, lower-cost service tier from premium strategic or storytelling work.
Q: Is it risky for a business to rely mostly on AI-generated ads?
A: The research suggests the main risk isn’t obviously bad AI output but mediocre “good enough” content that gets approved internally yet quietly underperforms in the market, so human review remains valuable.
Q: Will AI eventually get better at emotional storytelling in ads?
A: The study doesn’t rule that out, but as of this research, AI-generated creative consistently struggled whenever a brief demanded emotional nuance or a genuine point of view compared to human-made work.
Q: What’s the single biggest lesson for someone trying to build a career around AI content creation?
A: Position AI-assisted work as fast and scalable, and position your own strategic or emotional insight as the premium offering, since that’s where the data shows the clearest human advantage.
Q: Where can I read the original research?
A: It was published by Ipsos in partnership with Syracuse University’s Newhouse School of Public Communications, with additional coverage and analysis in Harvard Business Review.
