With the rise of AI, the way information is retrieved from the web is already shifting—from humans navigating websites to machines harvesting, compiling, and acting on information. What is discussed far less is what may follow: a shift not only in retrieval, but in how websites are designed, optimised, and even justified. This affects disciplines around UX, but actually goes much deeper.

For decades, the web has been built around graphical user interfaces—GUIs—designed for human–computer interaction. Making those interfaces work has always been expensive and complex. Entire disciplines emerged around this effort: usability, interaction design, user experience, service design, and many more. Over time, these practices proved their value, even if they constantly required organisational alignment, process changes, internal persuasion, and—not least—budget.

Today, much of that effort is quietly being reduced or removed. The reasons are manifold and not the focus here. What matters is an accompanying, more structural development: if AI agents increasingly act on behalf of humans when interacting with the web, the primary “user” of many websites may no longer be human at all. Why, then, continue to invest heavily in interfaces built for humans?

To some extent, we have already seen this before. Search engine optimisation made parts of the web machine-readable long before AI agents appeared. But SEO was always layered on top of a human-facing interface. Websites were still built to serve people first; machines came second.

That hierarchy may now be reversing.

If an AI agent knows its owner’s preferences, constraints, trade-offs, and goals far better than any website ever could, why should that website invest heavily in guiding, nudging, or persuading users through carefully designed interfaces? Why spend time, money, and organisational energy on UX when the actual interaction happens elsewhere—inside an agent that does not see, feel, or care about visual design?

This is no longer hypothetical. For many informational queries, human visits to websites are already declining. Users receive compiled answers, summaries, and recommendations without ever seeing the source. As human traffic drops, the cost per remaining human user rises—unless optimisation efforts are reduced. The direction of travel is hard to miss.

Consider a concrete example.

A large holiday booking portal has invested heavily in UX to guide different types of users toward conversion. Its main KPI is bookings, and its design logic is built around approximations: current search terms, click paths, inferred personas, acquisition channels. At best, it works with educated guesses.

Now contrast this with an AI agent searching for a holiday on behalf of its owner. The agent knows the person’s budget limits, travel history, preferences, sensitivities, and what “best value” actually means to them. When this agent “visits” the booking portal, it does not browse or explore. It queries, validates, and compares. It may not even need a “booking portal” at all—there will be something else. The carefully crafted interface becomes largely irrelevant.

This raises a deeper question, especially for large parts of the web whose business models still rely on the presentation layer. Much of today’s online information is financed indirectly through human attention—through advertisements embedded in human-readable interfaces. If humans no longer see those interfaces, the revenue logic collapses. What happens to the availability and quality of information when the economic foundation that funded its creation disappears?

Following this line of thought, it is not hard to imagine a future message that reads: No human-readable website available.

Not because information is gone, but because maintaining a dedicated human interface no longer pays off. We have seen similar shifts before: banks closing physical branches, Denmark removing post boxes, airlines serving only digital tickets.

In parallel, another effect may emerge: sameness. As interfaces themselves are increasingly generated by AI—cheap, fast, and good enough—many will converge toward similar patterns, either personalised on the fly or built from standardised templates learned from existing data. The web may start to resemble a phase familiar from other media transitions. At one point, music television was full of creative promise; later, most videos began to look alike—not because creativity vanished, but because optimisation logic converged. MTV recently closed its last remaining on-air appearance.

What do you predict?

Which services will still invest in refined, dedicated user interfaces? Which domains will continue to require humans to explore, feel, trust, and decide through interaction? And which will quietly skip the presentation layer altogether?