A.I. Ketchup: when the machine draws your bottle

Case study · Marketing 475 · Building Your Brand’s Visual Voice with AI

Heinz’s AI experiment worked because more than a century of consistent packaging had already taught the world, and the internet, what ketchup looks like.

The situation

Heinz has sold ketchup since 1876, when it was first sold as “catsup.” The “57 Varieties” line dates to 1896. Over the decades the brand kept returning to the same few visual cues: the glass bottle shape, the keystone-shaped label, the Heinz name and the number 57.

In early 2021, Heinz and its agency Rethink Canada ran a test called Draw Ketchup. People on five continents were asked to draw ketchup with no other instructions. Most of them drew Heinz: the name, the bottle shape, the keystone label and the 57. Heinz put the drawings on digital billboards in Canada.

The challenge

By 2022, Heinz was a brand more than 150 years old. In its awards entry, the agency framed the risk plainly: an established brand can lose relevance with younger people.

At the same moment, AI image generators were everywhere in the news and on social feeds. People were typing strange prompts and sharing the surreal results. Heinz needed a way into that conversation that was about the brand, not just about the technology.

What they did

In July 2022, Heinz and Rethink extended Draw Ketchup to a machine. They gave DALL-E 2 prompts that never named Heinz, such as “ketchup,” “ketchup art,” “ketchup renaissance” and “impressionist painting of a ketchup bottle.” Heinz said the results were almost always distinctly Heinz.

Heinz then turned the images into a campaign.

• A video montage of AI-generated ketchup images, plus a steady stream of images on Twitter and Instagram.
• An out-of-home activation in downtown Toronto.
• An open invitation for fans to suggest new prompts, with Heinz generating and posting some of them.
• According to the agency’s awards entry, a limited run of bottles with AI images in place of the standard label, and a virtual art gallery.

What happened

Jacqueline Chao, a senior brand manager at Heinz, summed up the message: even “the most unbiased source recognises that when it comes to ketchup, it has to be Heinz.” Trade press picked up the story within days.

The campaign went on to win awards in 2023.

• Silver Lion in Social and Creator at Cannes Lions 2023 (Cannes Lions).
• A 2023 Clio Gold (Ads of the World listing of the Clio entry).
• Over 850 million earned impressions worldwide and a social engagement rate 38% higher than past campaigns, as claimed in the agency’s awards entry. These figures come from the entry, not an independent audit.

Why it worked

Image generators learn from huge numbers of pictures. When a model draws “ketchup,” it draws whatever ketchup has looked like most often in those pictures. Heinz has a large share of the market (Muse by Clios reported about 60% in the U.S. and about 80% in Europe), and it has kept the same core look for generations. So the pictures the model learned from were full of Heinz bottles.

That is the real lesson. The AI did not make Heinz recognizable. Decades of consistent packaging did. The AI just made that hidden asset visible in a way people could share. The human insight came first: Heinz already knew from Draw Ketchup that people picture its bottle when they hear the word. AI gave it a new, timely way to prove the same point.

Where it gets complicated

This was a marketing demonstration, not a scientific test. Heinz and its agency chose the prompts, and the public saw the images they chose to share. Heinz’s own description was “almost always,” not always. Students should read the claim as persuasive, not proven.

AI image tools also have limits. They often garble text, so labels can look like Heinz without spelling Heinz. The effect depends on shape, color and layout, which is exactly why a stable visual system matters. It also means the trick only works for category leaders. A small brand that asks AI to draw its category will most likely get a picture of the leader. For most brands, this case is a warning as much as an inspiration: if AI does not picture you, your look may not be distinct or consistent enough yet.

The marketing principle

A visual identity is built through repetition, not through a single launch. Heinz kept the same bottle, label shape and cues for so long that people and machines learned them as the picture of the whole category. When AI tools are trained on what the world has seen, consistency becomes an advantage that shows up even in the machine’s output. Build a look strong enough that even AI recognizes it, and remember that the idea to prove it still came from people.

Sources

• Heinz Asks AI to Draw Ketchup, LBB Online (2022)
• Asked to ‘Draw Ketchup,’ Even A.I. Chose Heinz, Muse by Clios (2022)
• Heinz finds ketchup synonymous with its brand packaging in Canadian consumer experiment, Packaging Insights (2021)
• A.I. Ketchup, Ads of the World (2023)
• Heinz A.I. Ketchup (Social Insights & Engagement entry), Cannes Lions / Love The Work (2023)
• Heinz Company, Britannica (n.d.)

---

Back to Building Your Brand’s Visual Voice with AI · All case studies

Brand names, images and trademarks belong to their owners and appear here for educational purposes only.