Reaching Humans and Machines at the Same Time

For 70 years, the Cannes Lions have honored the world's best advertising. When a recent study by Jellyfish and INSEAD, conducted in partnership with Brandtech and presented in Cannes, asked AI to evaluate the same award-winning campaigns, the results showed virtually no alignment with human jury decisions. The correlation was zero.

The study suggests that humans and machines process and evaluate creativity and information in fundamentally different ways. Content that genuinely resonates with people is emotional, heuristic, and story-driven. For machines, that type of content is often difficult to interpret. Conversely, content created solely for LLMs – systematic, fact-focused, and ranking-oriented – often fails to engage human audiences.

To remain relevant to both audiences, brands need strong brand leadership. The same elements that matter to people are also reflected by LLMs as core brand attributes: long-term positioning, consistent branding, and distinctive assets. Much of a brand's visibility in AI systems is derived from these factors. However, LLMs are not human. They "read" creativity differently. Humans and machines perceive information differently because brains and processors work differently. This creates a fundamental tension for brand content, especially in e-commerce. Product pages that engage customers emotionally are often difficult for AI systems to interpret. Pages perfectly structured for machines can feel about as appealing to humans as a spreadsheet. Brands that continue to focus solely on emotional storytelling risk losing visibility in AI recommendations. Brands that optimize exclusively for AI risk losing customers before the purchase decision is made.

The good news is that this conflict can be resolved. There is no need to create two separate versions of a product page. Humans and AI can both be addressed effectively when content follows a structure that serves both audiences within the same piece.

What does this type of dual-optimized e-commerce content look like? The common denominator is specificity.

A product name, a numerical value, a genuine customer testimonial, a test video, or a before-and-after image can create credibility and emotional appeal for shoppers. At the same time, these elements provide AI systems with reliable, extractable facts and assessments that can be used in recommendations and price comparisons. By contrast, vague, adjective-heavy, or generic statements tend to persuade neither humans nor machines.

Returning to the backpack example: In an online store optimized for both humans and AI, each paragraph consists of two sequential elements:

• Element 1 – Answer (1 sentence):
  A direct, fact-based core statement. This is the layer that AI can extract and cite in recommendations.
• Element 2 – Evidence & Story (2–4 sentences):
  A concrete example, statistic, customer quote, or short anecdote that supports the claim. This is the layer that builds trust and anticipation while also providing AI with additional factual material it can reference.

For the same product attribute, the difference looks like this:

However, there is a caveat. According to the research team behind the Cannes study, overloading content with information risks creating "a second crisis of creative effectiveness" – a mistake already observed during periods when digital marketing became overly performance-driven.

The most important strategic shift is this: less generic volume, more original evidence and proof points. Content that an LLM could generate from the headline alone, without any brand-specific context, provides too few new signals and is therefore less likely to be cited or prioritized.

Tip 1

Build content as an "answer system" rather than as a continuous narrative. Every page should address one primary question, several supporting questions, and conclude with a clear, evidence-based takeaway. This makes content useful for readers and citable for LLMs. Remove unnecessary marketing language and generic introductions wherever possible.

Tip 2

Create a clear content structure:

• H1 headline focused on a clear problem statement or primary question
• H2 subheads that reflect real user questions
• Paragraphs that communicate one key idea at a time
• Specific entities instead of vague wording
• Lists for steps, criteria, and comparisons
• Short section summaries only when they provide genuine additional guidance

Tip 3

Integrate proprietary data, tests, observations, customer feedback, and reviews. Develop pages with FAQ, how-to, or article-based structures where appropriate. Use structured data to help machines recognize content types and entities more easily.

Speaking of backpacks: If you're considering buying a new one for your next fall hiking trip but are unsure about the ideal capacity, shoulder width, or material, helpful content can certainly be created to answer those questions. The fitness required for the mountains, however, is something neither AI nor the perfect backpack can provide.