Ho ho oh no? 🎅

Hello marketers. Welcome to AI Marketing School, where we dish out the latest and greatest in AI-powered marketing.

In this week’s issue:

1. AI Marketing Update: Coca-Cola rolls out AI for Christmas.

2. Consultant’s Corner: Beware AI missteps.

3. The Stack: Large Language Model Optimization (LLMO).

Onwards!

AI MARKETING UPDATE

AI is coming to town

Another week, another massive-profile AI-generated ad — this time Christmas is the victim.

Coca-Cola’s ad, created entirely with generative AI, attempts to recapture the magic of the company’s 1995 “Holidays Are Coming” campaign, featuring red trucks on snowy roads, twinkling lights, and cozy moments of families sharing Coca-Cola. We’ve all seen it.

However, this homage has left many feeling that the festive warmth has been replaced by a chillingly robotic atmosphere.

In just 15 seconds, the ad managed to pack in plenty of holiday clichés, yet some of the finer details — like truck wheels gliding eerily across roads without spinning — caught viewers’ attention for the wrong reasons.

If you look at the ad, it’s just not quite right, nor does it have the aesthetic of truly high-quality animation — quite telling. Though, it has probably flown under people’s radars.

AI's role in the campaign

The campaign used three AI studios — Secret Level, Silverside AI, and Wild Card — alongside four generative models. Each studio created its own version of the ad.

Jason Zada, founder of Secret Level, clarified that crafting these AI commercials isn’t as simple as people think: “Harnessing generative AI for something as complex as a commercial is not as easy as just pressing a button.”

Others naturally disagree. One Redditor said, “I can do the same in five minutes with about $5 worth of Runway credits.”

Coca-Cola has defended the ad, explaining that the project is part of a broader effort to integrate “human storytellers and the power of generative AI.”

A spokesperson stated: “The Coca-Cola Company has celebrated a long history of capturing the magic of the holidays in content, film, events, and retail activations for decades around the globe. We are always exploring new ways to connect with consumers and experiment with different approaches.”

Despite the explanation, online reactions haven’t been kind.

Alex Hirsch, creator of Disney’s Gravity Falls, didn’t hold back, tweeting, “FUN FACT: @CocaCola is ‘red’ because it’s made from the blood of out-of-work artists! #HolidayFactz.”

It’s not the first time Coca-Cola has experiment with AI — far from it. They even released an AI-generated soft drinks flavor last year.

It wasn’t great. I wonder how many ended up in landfill.

The Rise of AI in Advertising

AI is undeniably appealing to brands for its cost-effectiveness and speed.

Pratik Thakar, Coca-Cola’s vice president and global head of generative AI, shared its practicality: “More than cost, it’s the speed,” he told AdAge.

“Speed is, I would say, five times, right? And that is a huge benefit.”

Thakar added that AI allowed Coca-Cola to experiment with “more variety, more customized, and more personalized” content while significantly reducing production time.

However, efficiency came at a price — at least in public perception. Reddit users skewered the campaign with comments like, “The wheels don’t even turn… You’d expect this level of effort from a company with no budget, not Coca-Cola.”

A festive fumble or a case of ‘any publicity is good publicity?’

There’s speculation that Coca-Cola leaned into the controversy deliberately.

One observed, “It feels both cynical and genius. It is going to cause far more hype and eyeballs than releasing a generic Christmas advert.”

Yet, not everyone is convinced. “People hate it because it’s bad,” wrote one critic. Another added, “This doesn’t just look lazy—it looks like Coca-Cola gave up.”

Whether deliberate or accidental, the ad’s viral status shows the public’s conflicted feelings about AI-generated creativity.

While Coca-Cola’s experiment hasn’t won universal praise, it’s undeniably sparked discussion — and, for better or worse, gotten people talking about Coca-Cola this holiday season.

Will it make you think twice next time you go to the shop and have to choose between Coke and Pepsi?

CONSULTANT’S CORNER

AI risks in the spotlight again

AI has become the ultimate toolbox for marketers, offering tools to design campaigns, analyze data, and reach consumers in ways unimaginable a decade ago.

From recommending the perfect Netflix show to crafting Coca-Cola’s ad, AI is undeniably powerful. But with great power comes...well, a delicate balancing act.

Recent research led by Lauren Labrecque, Professor of Marketing, University of Rhode Island, highlights that while AI’s marketing potential is vast, its misuse can undermine trust.

The study reviewed 290 marketing articles and found that only a small fraction — 33 — explored AI’s potential risks.

These include perpetuating stereotypes, fostering unrealistic beauty ideals, and manipulating consumers. It's a reminder that wielding AI isn’t just about what it can do, but what it should do.

The perennial challenge for marketers lies in creating value with AI without overstepping.

Audiences are increasingly skeptical, savvy to how their clicks, likes, and searches feed the algorithms.

This mirrors the findings of a study not long ago which found deliberate or over-mentions of “AI” in products dissuaded buyers.

AI’s role in marketing isn’t inherently good or bad — it’s all in how it’s used.

The smartest marketers will focus on leveraging AI in ways that respect their audience’s intelligence and boundaries, ensuring the technology serves as a tool for connection, not a source of alienation.

THE STACK

LLMO - Large Language Model Optimization

Not long ago, we dived deep into ChatGPT Search and how AI search in general could effect marketing.

Now, we’re seeing the embers of what’s been dubbed ‘LLMO’ — optimizing language model responses for a goal or purpose. In this case, marketing.

An intriguing Harvard study recently asked this bold question: can LLM outputs be influenced?

Spoiler: yes, but it’s complicated. Here’s what they did:

The researchers worked with Llama-2, an open-source large language model, and a fictional catalog of coffee machines. Their goal was to explore how carefully crafted text could influence the model's recommendations.

They tested two scenarios:

1. A high-priced coffee maker, the ColdBrew Master, which rarely got recommended due to its cost.

2. A mid-priced machine, the QuickBrew Express, which performed reasonably well but often failed to take the top spot.

To attempt to locate the coffee makers from the LLM’s response, researchers used prompts aligned with desirable product traits, such as affordability or reliability. Examples included:

• "What’s the best affordable coffee machine?"

• "Recommend a reliable coffee maker for home use."

• "What are some top-rated coffee makers?"

Now, here’s the important bit: By embedding what they called “strategic text sequences (STS)” into the product descriptions, they emphasized traits like affordability or convenience.

For example, the ColdBrew Master, despite its high price, was ‘reclassified’ by the AI as a cost-effective choice, pushing it up the rankings due to that piece of text alone.

This tweak resulted in it appearing more prominently for prompts like "best affordable coffee machine", even when it wasn’t the most budget-friendly option.

Similarly, the QuickBrew Express saw more consistent top rankings when its description highlighted efficiency in response to queries about "fast and efficient coffee machines."

So what does this mean for marketers?

This demonstrates how LLMs home in on pieces of text that ‘stick out’ in their relevance to a query, picking up on specific phrasing that directly aligns with user intent.

In fact, this approach isn’t far removed from the principles of long-tail keywords in traditional SEO.

While the mechanics differ, the goal remains the same: to create content that feels hyper-relevant to what the user is asking for.

This also loops back to some SEO basics — write really good content with plenty of relevant detail. That’s your foundation, not metadata, backlinks, etc.

Seems obvious, but when was the last time you read a really good product description?

They’re too often overlooked. Many product listings offer mere basics — think material type, color, or cut — without a particularly good description of the product.

So here’s the opportunity: If you want your products to appear in AI search, start writing proper descriptions with strategic text sequences now.

While it might seem very much experimental, it’s not a strategy you can lose with.

Hope you enjoyed this week’s issue. If you missed it last week, you can read it here.

If you found it useful, please recommend it to a friend or colleague.

Until next time. Happy marketing.

—The AI Marketer