top of page
Search

When Marketing Looks “Off”

Updated: Oct 2

AI-generated ads promise speed and creativity, but when visuals feel “off,” brands risk losing consumer trust. This article explores how the uncanny quality of AI images can backfire—and why authenticity still matters more than perfection.


ree

It was just after 6 p.m. in Philadelphia. The sidewalks were still warm from the day’s sun, and the city hummed with the sound of workers making their way home. One man, a mid-level office employee, had his usual route—same streets, same corner store for a soda. But something unusual caught his eye. A team of workers was pasting up a new poster across the brick wall outside a laundromat. It was a marketing ad showed a smiling woman holding a bottle of surface cleaner, surrounded by a surreal, hyper-clean kitchen. Something about her hands didn’t sit right. Her eyes seemed just a little too symmetrical. The lighting? Unnaturally even. It looked fake—because it was. Clearly AI-generated.


The man paused. He frowned. It wasn’t anger—just something closer to disappointment. He didn’t know the brand well, but now he didn’t want to. The image felt off, hollow. A brand trying to cut corners rather than connect. Days later, as he browsed cleaning products at the store, he reached for a different bottle.


This quiet, mundane moment encapsulates a growing issue in the marketing world: consumers are increasingly sensitive to the feel of authenticity, even when they can’t quite identify the technical source of their discomfort. And when AI-generated images fail to meet emotional and visual expectations, brand trust suffers.


Despite advances in generative AI, consumer trust lags behind its capabilities. A study by PMC (2025) revealed that nearly 40% of AI-generated images were mistaken for human-made yet viewers still rated human-created visuals as more credible and realistic. It’s not always about detection. It’s about perception. The human brain picks up on subtle cues: the texture of skin, the way light falls, the inconsistencies that suggest reality. AI images, especially when generated from vague or overloaded prompts, often miss these details—producing what researchers call the "uncanny valley" effect.


Consumer attitudes are also deeply split. A Zappi (2025) survey found that about one-third of respondents liked AI in advertising, another third felt neutral, and the remaining third disliked it. Men and younger users tended to be more accepting, while older demographics and women were more skeptical. Most notably, 30% of participants reported that AI-generated ads made them like a brand less.


Marketing missteps using AI aren't hypothetical—they're happening. Coca-Cola’s 2024 Christmas campaign, for example, tried to revive a beloved 1995 ad using AI-generated visuals. The result? Eerie facial expressions, trucks gliding without turning wheels, and distortions that were more horror than holiday. Though the company produced dozens of versions, most were scrapped due to quality issues. What survived still damaged the brand’s reputation, proving that AI shortcuts come with real costs.


These failures often stem from poor prompt engineering—either too vague or overly complex. Common issues include distorted anatomy (extra limbs or fingers), mismatched lighting, or generic, lifeless environments. Failing to use negative prompts—keywords that prevent unwanted features—leads to visual errors that viewers notice, even subconsciously.


Not every AI campaign ends in disaster. When approached with care and intentionality, AI-generated visuals can enhance marketing impact. Unigloves’ 2025 “Derma Shield” campaign is a strong example. They produced over 250 photorealistic images across professions using Midjourney and Adobe Firefly, refining them further in Photoshop to ensure natural skin tones and lighting (Madgicx, 2025). Here, AI was a starting point, not the final product.


Puma’s 2025 AI-driven ad took a narrative-first approach. Instead of focusing on the technology, it told an inclusive, emotionally resonant story. A follow-up study by Zappi (2025) revealed that most viewers remembered the story and message, not the fact that the visuals were AI-generated. This illustrates a key principle: strong storytelling can mask, and even enhance, AI-generated content when executed effectively.


Behind every great AI image is a well-written prompt. Effective prompts are detailed but not overloaded. They include elements like subject, environment, camera settings, lighting, texture, and quality cues (“DSLR,” “RAW photo,” “soft sunlight”). The most advanced campaigns also used negative prompts to eliminate unwanted features like blurriness or unnatural anatomy.


Iterating on prompts—refining them based on output—is also essential. Successful brands often combined automated generation with human editing, ensuring that images didn’t just look realistic but felt authentic.


AI image generation isn’t going away. But it must be used responsibly. Brands need to consider their audience, refine their tools, and always put story and authenticity first. A striking image means nothing if it lacks the emotional resonance of real human connection.


The gentleman in Philadelphia didn’t need to know about prompt engineering or visual artefacts to feel something was off. He simply saw an image that lacked heart. And in that moment, a brand lost him—not because it used AI, but because it forgot the human behind the sale.


By: Tom Foreman, Founder




 
 
 

Comments

bottom of page