Description

In this takeaway piece, Abi reflects on Slobodan Manić’s Experimentation Elite keynote and the uncomfortable truth sitting underneath it: the basics were never basic. Semantic HTML, accessible interactions, clear labels, robust forms, meaningful content and usable structure were always load-bearing. The human case was already enough. AI has just made the consequences of ignoring those basics more visible.

The piece starts from a familiar digital crime scene: divs pretending to be buttons, labels that do not label, important content loaded late, vague claims with no proof, and forms that fail silently. Humans have learned to work around these broken systems by guessing, retrying, rage-clicking or quietly abandoning. AI agents are less forgiving. They rely on structure, labels, clear states and extractable information. Where humans improvise, machines often fail.

Abi pulls out a practical four-part lens from Sani’s keynote: identity, structure, content and interaction. Can a machine tell who you are? Can it extract the right information? Can it rely on what you say? Can it actually use the thing? These are not mystical “design for bots” questions. They are the same questions good UX, accessibility and content work should already have been asking.

The strongest point is also the most irritating one: accessibility, semantics, structure and performance helped humans all along, but now AI needs them, so suddenly everyone has remembered they matter. Useful? Yes. Infuriating? Also yes. The conclusion is blunt: AI did not raise the standard. It exposed who was already below it.

Topics

• Accessibility

• Semantic HTML

• AI agents

• Machine-readable websites

• Structured content

• UX foundations

• Robust interaction design

• Forms and labels

• JavaScript-heavy websites

• Content clarity

• Identity consistency

• Human-centred AI readiness

• Agentic browsing

• Assistive technology

• Upstream optimisation

• Meaning as infrastructure

Best for

UX designers, developers, accessibility specialists, CRO practitioners, product teams, content designers, SEO teams, AI-readiness teams, and anyone currently discovering that their beautiful JavaScript soup is about as machine-readable as a ransom note in a blender.

Background

This piece was created from Abi’s main takeaway after Slobodan Manić’s Experimentation Elite keynote. It fits the Key Takeaways category because it translates a conference talk into a practical industry reflection for people working across UX, experimentation, accessibility, AI and digital strategy.

It also connects strongly to Abi’s wider work on usability, interface quality and upstream optimisation. The argument is not that websites should be designed for bots instead of people. It is that the same foundations that help humans also help machines: clear identity, meaningful structure, reliable content and usable interaction.

From a Corpus perspective, this sits right on the seam between Interface and Model. The page may no longer be the destination, but it is often still the source material. If the interface is structurally weak, semantically vague or dependent on decorative nonsense, the model has less to work with. Meaning is infrastructure. And, annoyingly, it always was.