Machine Readers and Brand Intention: When Algorithms Shape Perception

The Machine in the Middle

When schema.org launched in 2011, it did something quietly consequential: it created a formal grammar for machines to read brand content and surface meaning before any human clicked through. A structured data tag on a recipe page, a product listing, an event announcement — these weren't design decisions in any traditional sense, but they shaped what appeared, how, and to whom. The human experience of a brand began, in a small but measurable way, downstream of a machine interpretation. Most brand teams didn't notice, or didn't need to. The stakes were low enough that it felt like an SEO problem, not a brand problem.

At an AirOps Next panel, James Hurst participated in a conversation about how craft, culture, systems, and storytelling evolve when machines are part of the audience — not as tools that execute human intent, but as readers, curators, and increasingly as co-authors. The framing was careful and genuinely curious. But it left something unresolved, and that unresolved thing seems worth sitting with.

The working belief James returns to across his writing and talks is this: the product builds the brand. The actual experience of the thing itself is what convinces — not the campaign, not the identity system, not the positioning document. It's a principle that has held up well across a long career at companies where the product and the brand were, at their best, nearly indistinguishable. But it was a principle formed in an era when the path from brand intention to human perception was relatively direct. Something was made. Someone encountered it. An impression formed.

What happens to that principle when something sits between the making and the encountering?

Spotify's Discover Weekly, launched in 2015, is one of the clearest documented cases of craft bending toward a non-human reader. Artists and labels began adjusting track length, intro timing, and sonic texture — not because listeners asked for it, but because the algorithm rewarded certain patterns and penalised others. The human audience was still the destination, but the machine was the gatekeeper, and creators started designing for the gate. Whether that's a corruption of craft or simply a new constraint — the way radio edits were once a constraint — is genuinely unclear. But it marks something: the moment when a non-human reader became a design brief.

Adobe's Firefly integration into Creative Cloud, positioned in 2023 around AI as a collaborator rather than a shortcut, pushed this further into the authorship question. If a brand's visual language is generated, curated, and refined through a system trained on vast corpora of existing visual culture, what does it mean to say the brand has a voice? The craft is still happening. The curation is still human, or partly human. But the loop between intention and output has a new participant, and that participant has aesthetic preferences baked into its training.

James' book Weird the Normal / Normal the Weird argues that the real discipline isn't landing on strangeness or on structure — it's the switching between them. Weirding the normal to expose the assumption underneath; normalising the weird to give a strange idea enough structure to survive. The book describes this as a pulse, a way of staying alive. What's interesting, placed next to the machine-audience question, is that algorithmic systems are extraordinarily good at one half of that move and structurally resistant to the other.

Algorithms normalise. They find patterns, reward consistency, surface what resembles what already worked. They are, in a functional sense, engines of the familiar. The weirding move — the deliberate estrangement, the tilt that makes something visible again — tends to get filtered out before it reaches a human eye. Novelty without prior signal looks, to a machine reader, like noise. The system isn't hostile to novelty; it's indifferent to it.

So here is the question the AirOps Next panel approached but didn't quite resolve: if machine readers are now shaping perception upstream of human experience, and if those readers have a structural preference for the legible over the strange, does the principle that the product experience is the brand still hold — or has something changed about what counts as the product?

One possibility is that the brand now has two products: the human-facing experience that James' framework has always been built around, and the machine-readable signal that precedes it. If that's true, then the design question isn't only what does this feel like to encounter? but also what does this look like to a system that has never encountered anything? Those are different questions. They may require different answers. And it's not obvious that the same team, with the same instincts, is well-placed to hold both.

Another possibility is that the principle holds, but the definition of experience needs expanding. If a machine interprets a brand's structured data, surfaces it in a particular context, and shapes the expectation a human arrives with — is that not already part of the experience? The moment of encounter has simply moved earlier, and become less visible.

What would it mean to design for that earlier moment without losing the thing that makes the later moment matter?