AI slop costs a brand more than it saves. Lessons from McDonald’s, Dafi and Kraft Heinz

Three days of a McDonald’s Christmas ad

On 6 December 2025, McDonald’s Netherlands posted a 45-second Christmas ad on YouTube made from generative AI clips. The comments were merciless, and the most common complaint was that the AI was obvious. As BBC News reported, three days later the company took the video down. In a statement it said the episode was an important learning as it explored the effective use of AI.

The most telling part came afterwards from Melanie Bridge, CEO of The Sweetshop, the studio that made the film with the agency TBWA\Neboko. She defended the work: seven weeks, thousands of takes, an edit like any high-craft production. We believe it. But viewers weren’t judging the effort. They judged the result and how well it fit the brand.

Dafi and the spot that vanished

A few days earlier, a Polish brand learned a similar lesson. As NowyMarketing reported, Dafi, a maker of water filter jugs and bottles from Poland’s Świętokrzyskie Mountains, posted a spot on social media that was entirely AI-generated, with no mention of that anywhere. The post stressed that the company had been making its products in the heart of the mountains for more than 30 years, and people associate the brand with nature and clean water. The artificial look clashed so badly with that image that negative comments piled up and the video quickly disappeared.

To add irony, in an earlier campaign with the Polish singer Doda the brand had argued that you don’t need artificial intelligence to think. Antyweb wrote about a nationwide storm caused by one badly used tool. Dafi didn’t comment publicly, so we rely on media reports.

What the data says

These aren’t isolated cases. In a Canva study run with The Harris Poll (1,415 marketing leaders and 3,547 consumers in seven countries, published in May 2026), 41% of marketing leaders called AI slop a considerable challenge. Meltwater data cited in the report shows a ninefold rise in mentions of “AI slop”. 78% of consumers would rather see ads made by people, even if AI could make them better, and seven in ten say AI ads are missing their soul.

The most useful finding comes from a DoubleVerify study from July 2026 covering 22,000 consumers in 22 markets. 42% would think worse of a brand because of low-quality or “uncanny” AI advertising, yet 40% view polished AI ads positively, and 56% admit they can’t consistently tell AI content apart. Customers don’t punish AI as such. They punish sloppiness and content that doesn’t fit the brand.

Kraft Heinz built the brand into the tool

Kraft Heinz is a good example of AI that doesn’t produce slop. With its partner Apply Digital, the company built a platform called TasteMaker on Google Cloud. According to a Google Cloud case study, the models generating text, images and video draw on what the company calls its brand intelligence: guidelines, product catalogue data, photo libraries and more, gathered in one place. Instead of an empty chat window, teams get ready-made apps for specific tasks.

“The first was to bring our brand intelligence into the platform.”

The results Google Cloud reports: content for new products takes 8 hours instead of 8 weeks, and 70% of the product development and marketing users who have access use the platform. These are the technology vendor’s figures, so we treat them with care. The direction is clear, though: the brand is part of the tool, not a line added to a prompt.

L’Oréal: rules first, models second

L’Oréal took a similar path and added firm rules. In its internal CREAITECH lab, marketing teams use Google’s models to create concepts, storyboards, packaging redesigns and test product shots in different settings. According to Retail Dive, the time to produce initial concepts dropped from weeks to days, with lower costs.

What L’Oréal doesn’t do with AI matters even more. As Google Cloud describes it, the group doesn’t generate people for advertising and doesn’t use life-like AI faces, bodies, hair or skin to show product benefits. Since 2021 it has had its own framework for responsible AI, seven principles overseen by internal bodies. The limits are set before anyone opens a tool.

What this means for your brand

What McDonald’s and Dafi were missing was anything between generating the content and publishing it that looked after the brand. Kraft Heinz and L’Oréal also use AI at scale, but the brand is built into their process: guidelines and product photos are available to the tool, tasks are specific, limits are written down, and a person looks at the result before it goes out.

A smaller company doesn’t need its own platform to apply the same principle. It needs a brand described so AI can reproduce it, templates for the most common materials and a clear rule about who signs content off before publishing. We go into this in our post on teaching AI your brand. In our service this is the AI that knows your brand area.