The Challenge of Balance AI and Ethics In Marketing

Have you started to question what’s real and what’s not whenever you’re scrolling for content online? Yeah, me too. When I see something online, and my brain categorizes it as something made for humans and then I find out it’s AI, I think that we’re getting closer to the scenario depicted in the sci-fi novel series I’m reading, Hyperion Cantos, in which AI develops a mind of its own, emancipates from humanity and plots against us. AI tools have permeated every aspect of our daily life, including, of course, marketing.

On the bright side, brands now have a deeper knowledge of what their audience needs or wants to see and can adjust their strategy in real time. On the darker side, there’s a lack of transparency, consumer manipulation and deception, privacy issues and crappy, robotic content. We’re going to explore the challenge of balancing the use of AI to improve our marketing without crossing the line of manipulation and exploitation.

AI-powered marketing tools depend entirely on accessing and analyzing consumer data to provide the accurate insights we need to adapt tactics based on their real-time behaviour, allowing us to create those ultra-personalized marketing strategies we need to offer our audience the solution they’re looking for. Where are these tools pulling all that data? Of course, from users’ consent when they visit a business website, we are all very familiar with. One of the biggest challenges is ensuring that users really understand what they’re agreeing to when they share their information. And even with their consent, things can get opaque from here, as users (almost) never read the cookie and privacy consent banners most websites display.

Depending on the nature of your business, you could be dealing with private and sensitive data like credit card and bank details, or identity information that can put even well-intentioned companies in a delicate position, as they can unknowingly cross boundaries by using data in ways that consumers didn’t anticipate. Data is also the main target of security breaches, and having your customers’ data leaked online is a nightmare that can severely damage your business reputation, especially if yours is a small business, often considered “easy targets” by hackers.

For marketers, this scenario creates a paradox. While we need as many details as possible, we also need to ensure we obtain meaningful consent and comply with the data protection laws (some notable ones include GDPR in Europe, CCPA in California and PIPEDA in Canada), and we ensure the data we have is safe. Seems like a lot, but it’s possible. Here are a few ideas:

• Use Clear Language: While no one is going to read a consent banner with a long legal paragraph, using simple phrases like “We use your data to offer you a better experience,” “We’re committed to protecting your information,” or “We will never sell your data” will do the trick.

• Add Timely Notices: Depending on the nature of your business, you may need specific pieces of information as users navigate your website. For example, if you’re an e-commerce site, adding a pop-up saying “We use your browsing history to recommend products” while the users are scrolling through products, or one notifying “We need your location to show the deals at stores near you” right when they start searching for those stores’ addresses. Requesting permission at the moment you want to collect specific data is far more effective than a one-time, heavy-text banner.

• Use Layered Consent Forms: In case one of the few people who want to read all the consent disclosure details enters your website, and to comply with transparency without overwhelming people, offer the option to expand for more information after the concise summary version.

• Offer Granular Control: Transparency also means letting people choose what they want to share.

While data transparency is essential to AI responsibly, it’s just a matter of remembering that you’d like to avoid a lawsuit to be on top of privacy compliance, so it isn’t the most worrisome aspect of AI use for marketing. However, the next ones are.

Going back to how AI is portrayed in Hyperion, humans don’t know how it works or where AI “lives” after becoming independent. The irony of a human invention being out of comprehension for its makers is not 100% sci-fi; something eerily similar is happening with real-world AI.

The “Black Box” problem refers to the inability of humans to comprehend or justify AI’s decisions, especially tools using deep learning, one of the most modern AI forms that uses multilayered neural networks to solve complex problems and make decisions in a similar way humans do. Since deep learning systems are trained in a similar way we learn in school, that is, learning by examples and association, until eventually we integrate that information, and we eventually forget how we learned it. A good example of this is when parents have to help their children with homework: They know how to multiply, but must fail to explain how it’s done.

University of Michigan-Dearborn’s Associate Professor Samir Rawashdeh explains that this is a big deal because, since developers don’t understand how an AI reaches a conclusion and they know it’s an error, fixing the issue that caused it becomes a challenge. In addition, ethics are put to the test, especially when companies delegate sensitive decisions to AI, such as health insurance companies leaving the decision of which patients get a treatment covered or not to algorithms.

While marketers don’t have someone else’s life in our hands, most of us are not AI developers, so trying to fix an error is out of the question. However, we can have a hard time justifying the implementation of some of the conclusions AI-marketing tools reach for tasks like targeted ads that can feel disturbing for consumers due to their level of accuracy and personalization, making them feel unsafe and stalked by the brands, especially when people start to see ads based on sensitive life events. Let’s look at one real-world example.

In 2012, Target used a predictive algorithm to identify customers who were likely pregnant based on changes in their buying behaviour as their pregnancy went on, such as starting to buy supplements during the first 10 weeks, fragrance-free lotion around the beginning of the second trimester, or large quantities of cotton balls and more fragrance-free hygiene products closer to their delivery date. The source of information came from the clients’ Guest ID number, tied to their payment information, purchasing history, name, or email address, which becomes a sort of library that has a record of everything they’ve bought and any demographic information Target has collected from them or bought from other sources. They also took into consideration whether those clients had registered with Target baby registries before.

Based on the algorithm results, they scored the clients to select the ones with higher chances of being expecting a baby, and they started to send these people discount coupons for baby products, including a teenage girl who hadn’t told their family the big news. Seeing that her daughter got maternity clothing coupons, her father went to a Target to yell at the manager, alleging that they were trying to encourage her to get pregnant. A few days later, a customer service employee called the dad to apologize again, only to find out the AI was right: His daughter confessed she was indeed pregnant.

Soon enough, they found out that many of the clients were creeped out by the fact that a corporation knew they were pregnant before announcing it to friends and family, so they became more subtle about it, sending the baby items coupons mixed into random ones like discounted beer. “... We found out that as long as a pregnant woman thinks she hasn’t been spied on, she’ll use the coupons,” said Andrew Pole, the statistician behind Target’s pregnancy prediction tool, to the New York Times.

This happened in 2012, and it’s one of the most tame examples out there. Imagine how much AI marketing tools can find out about us now.

Disturbing accuracy and consumer manipulation are tied together, but a (very) thin line still separates them. From psychological exploitation, granular targeting of vulnerable groups, shady pricing strategies and mimicking human authenticity and expressions, brands are now able to predict their buyers’ behaviour based on details as personal as their sleeping patterns to show them the products AI “knows” they’ll look for before they even know it themselves.

The following cases of companies’ algorithms crossing the lines of ethics and morality for the sake of money and control are coming to light more frequently:

Meta’s ad platform has always been praised by its robust and granular segmentation features, which allow brands to deliver ultra-personalized ads to people showing strong interest in their niche, but that comes with a dark side. In April 2025, former Facebook executive and whistleblower Sarah Wynn-Williams told a US Senate subcommittee that Meta actively targeted teens with ads when the algorithm identified they were sad, depressed, or unsatisfied with their body image based on their activity and interaction on Instagram and Facebook, by sharing that data with weight loss and cosmetics companies.

According to her, Meta considers the data of impressionable and mentally vulnerable 13 to 17-year-olds as “very valuable” for advertisers. “Advertisers understand that when people don’t feel good about themselves, it’s often a good time to pitch a product — people are more likely to buy something,” said Wynn-Williams during her testimony. Meta, of course, has denied these allegations and reiterated its commitment to reinforce advertisers’ compliance with its privacy policy guidelines. However, the company has a history of user data and privacy controversies and has faced several legal battles and fines worldwide because of it.

Adults are also targets of Meta’s algorithm. A 2021 Research study creating two test accounts, one for a 14-year-old and the other for a 29-year-old, has shown that whenever the accounts interacted with weight loss content, the platform starts to recommend weight loss accounts and to deliver ads accordingly, putting users in a loop of content reinforcing poor body image. Meta has acknowledged the loopholes in their security policies. However, unfortunately, it’s still happening: A confidential study run by Meta themselves revealed that among 1,149 teenagers, the 223 showing dissatisfaction with their appearance saw more than triple the amount of harmful content in their feeds. It also highlighted the flaws in the moderation tools, which failed to block 98.5% of the harmful content.

Now, the problem isn’t personalization or convincing people to buy a weight-loss tea; it’s the exploitation of insecurities to amplify them, manipulate vulnerable users and sell them products. Yes, all forms of marketing involve some form of persuasion, and AI can help brands to reach the people they want to persuade with the content they need. Still, persuasion and manipulation are interchangeable in the eyes of companies like Meta and marketers using AI can cross at any time when they forget moral principles for the sake of profit.

If your social media accounts have been flooded with content (marketing-related or not) that looks suspiciously uncanny valley, cheap, soulless and generic, you’re not alone. These pieces of robotic content, known as “AI slop”, have proliferated since the Pandemic. Despite big corporations’ hype, like Coca-Cola and Mondelez, that are pushing AI’s capabilities to escalate their marketing beyond human production limitations, consumers aren’t so hyped about it.

AI slop’s adverse sentiment, in marketing media specifically, has already been studied. A study conducted by NielsenIQ, using tools like surveys, implicit response time tests and eye tracking technology, revealed that even the most realistic and “flawless” AI-generated ads cause a degree of dissonance among consumers in both unconscious and conscious levels.

In other words, our brains have a hard time connecting with AI content, distracting us from its intention, as we wonder, “Is this real or AI?”, making it less memorable. The study also showed that AI ads are perceived as “annoying” or “boring”, which may even contribute to a negative perception of the brand as a whole.

When the uncanny valley effect is blatant, and it comes from a big company, people are vocal about their sentiment, as the following examples show:

Toy retail brand Toys “R” Us has been working to make its comeback after filing for bankruptcy in 2017, and they also wanted to present to new generations their legacy in the toy industry by sharing the company’s history and evoking the nostalgic feeling of toy stores feeling like magical places. That’s why they decided to make a new brand film telling the origin story of its founder, Charles Lazarus, and the birth of the brand’s mascot, Geoffrey the Giraffe.

As they wanted to be perceived as a modern company embracing changes and new technologies (and cut costs), they chose to use Sora, Open AI’s text-to-video generative-AI tool, to make the brand film using detailed text prompts describing scene, characters, setting, and actions, with minimal human intervention to apply “corrective VFX” (touch-ups, visual fixes) to polish and fix inconsistencies in the Sora-generated shots.

The film, named “The Origin of Toys ‘R’ Us”, became the first major brand film publicly produced with Sora. And while some praised it for its “innovation” and “boldness” for its use of cutting-edge storytelling technology, it had an overall negative reception among the public and creative professionals. The criticism focused on the oddness of it all: The “uncanny valley” effects, the awkward movements, and a sense of artificiality or “soullessness” in the final product.

Whether you celebrate Christmas or not, you probably associate this holiday with Coca-Cola’s ads and their characterization of the Santa Claus figure that has left a mark on many generations of children. Unfortunately, the magical and nostalgic aura of Coca-Cola’s Christmas is dead, and generative AI killed it.

The company started to use AI to create its Christmas campaigns in 2024. That first wave of AI-generated ads, which brought three AI studios together, was released in November and aimed to re-imagine the classic “Holidays Are Coming” ads from the 90s.

The campaign was received with harsh criticism from the public, calling it “uncanny,” “soulless,” and “emotionally hollow.” Others commented on the fact that the ads were made out of something created by real workers. The Gravity Falls creator, Alex Hirsch, said that the ad had so much red because it was made “from the blood of out-of-work artists.” It also highlighted the long road ahead of AI technology to replicate human-crafted, iconic pieces of media like the 90s ad, as what we saw was the best they could make out of many attempts of AI footage that was discarded.

Despite the 2024 failure, Coca-Cola hasn’t given up on cheap out and “optimizing” their marketing costs with AI, because they graced us with their 2025 AI Holiday campaign too, and they outdid themselves by making it worse:

This year, they didn’t use close shots of the “people” smiling with dead stares, waiting for the Coca-Cola trucks signaling Christmas arrival. Instead, they featured a variety of eerily-looking animals making expressions of surprise, joy and curiosity as the trucks passed by.

Will Coca-Cola moderate or stop its AI use? No. “We need to keep moving forward and pushing the envelope… The genie is out of the bottle, and you’re not going to put it back in.” Said Global VP and head of generative AI, Pratik Thakar, in response to this year’s ad backlash. They also used a team of 20 people this time; the team behind the 2024 ad was 50, so there’s more blood of out-of-work humans used to create this AI piece. How many people will be involved for next year’s ad?

While AI is now an integral part of how we do marketing and will continue to evolve, consumers still want human values. In a world where stress, isolation and emotional fatigue are prevalent, people want to escape for a moment from their hectic reality to a place where things are simple and can connect with the things that matter the most to them. Taking Coca-Cola as an example, their new holiday ads are so rejected because people don’t want to see human emotions replicated by a computer program; no matter how realistic it may look, it doesn’t have the spark the human heart has.

Robert Rose from the Content Marketing Institute highlighted the need for businesses to stop putting their focus on using the most advanced tools to maximize performance metrics and shift their attention to connecting to people. He said that the marketing planning for 2026 should be centred around three “Rs”:

• Resonance: Make marketing to connect with your audience, not just to reach them.

• Realness: Put human craftsmanship and creativity at the forefront of your strategy.

• Relationships: Nurture the connection you have with your audience with trust.

AI is a powerful resource that unlocked the potential for data we couldn’t access before, and that is why I’d like to add another R for Responsibility. Businesses now have access to the most detailed information about their consumers, but not all should be used to sell. Brands should handle people’s data the way they’d like theirs to be handled. Do you want a weight loss company to exploit your body image concerns to sell you a magic pill?

AI is not the enemy, and I don’t see it becoming something like the sci-fi novel I mentioned earlier, at least not in our lifetime. We should use it as an ally to enhance our human efforts in marketing and other areas of our lives.

Share