Imagine opening your analytics dashboard to discover something strange: your website is getting thousands of visits, but nobody is clicking, scrolling, or hovering over anything. The session durations are impossibly fast—entire workflows completed in milliseconds. Your carefully crafted call-to-action buttons? Ignored. Your beautiful hero images? Never viewed.
Welcome to 2025, where a recent study found that 63% of websites are now receiving visits from AI agents. Your new users don't care about your color palette, your carefully chosen fonts, or that perfectly placed "above the fold" content. They're here to extract data, complete transactions, and move on—all without a single human ever looking at your interface.
This isn't science fiction. It's the new reality of interface design, and it's forcing us to rethink everything we thought we knew about UX.
Welcome to 2025, where a recent study found that 63% of websites are now receiving visits from AI agents. Your new users don't care about your color palette, your carefully chosen fonts, or that perfectly placed "above the fold" content. They're here to extract data, complete transactions, and move on—all without a single human ever looking at your interface.
This isn't science fiction. It's the new reality of interface design, and it's forcing us to rethink everything we thought we knew about UX.
The "No More UI" Prediction
In early 2025, UX legend Jakob Nielsen published an essay that sent shockwaves through the design community: "Hello AI Agents: Goodbye UI Design, RIP Accessibility." His prediction was stark—by 2027 to 2030, websites may never see human users again. Instead, AI agents will handle all our digital interactions, negotiating with other systems on our behalf.
Nielsen's argument is surprisingly logical: if intelligent agents can understand our intent and complete tasks automatically, why would we waste time navigating websites ourselves? Your AI assistant could book your vacation, manage your finances, and order your groceries—all while you sleep. The traditional user interface, optimized for human eyes and fingers, becomes irrelevant when the "user" is an algorithm.
The backlash was swift. Critics published rebuttals with titles like "Jakob Nielsen Just Declared UI Design Dead. He's Spectacularly Wrong." But whether you agree with Nielsen's timeline or not, the underlying trend is undeniable: AI agents are already here, and they're interacting with our interfaces in ways we never designed for.
From UX to AX: A New Design Discipline Emerges
While designers debated Nielsen's predictions, a quieter revolution was taking place. In 2025, Mathias Biilmann, CEO of Netlify, coined a new term: Agent Experience, or AX. He defined it as "the holistic experience AI agents will have as the user of a product or platform."
This wasn't just clever branding. By March 2025, Netlify had hired Sean Roberts as the industry's first "Head of AX Architecture"—an executive-level position dedicated entirely to ensuring AI agents could effectively use their platform. ChatGPT now deploys over 1,000 sites daily to Netlify, and none of those deployments involve a human clicking through a visual interface.
Major corporations followed suit rapidly. Salesforce published guidelines on Agent Experience design in April 2025. Job postings for "AI Agent Experience Designers" and "AI Interaction Designers" appeared at companies ranging from Samsung to Adobe, with salaries ranging from $95,000 to $140,000. The message was clear: designing for AI users isn't a future concern—it's a present-day necessity.
The numbers back this up. According to a 2025 survey of over 5,700 developers, 24.3% are already designing APIs specifically with AI agents in mind. This represents a fundamental shift happening right now, not a distant possibility.
What Happens When Nobody's Looking?
So what does design actually look like when your primary user has no eyes? The answer is both simpler and more complex than you might think.
Think of it this way: you're no longer designing a magazine layout; you're designing a filing system. Every piece of information needs to be clearly labelled, consistently formatted, and logically organized. An AI doesn't appreciate beautiful design, but it definitely notices when your form field is named "input_37" instead of "customer_email."
One colleague learned this lesson the hard way while trying to automate a simple data download from a government website. The interface was unpredictable—buttons shifted position with each search result, fields lacked clear labels, and the workflow kept changing. What should have been a five-minute automation task became days of frustration as scripts kept clicking the wrong elements and entering data in the wrong fields.
This is sometimes called "Zero UI"—the idea that the best interface might be no visible interface at all. While the term itself has become controversial since being coined in 2015, the underlying technologies have gone mainstream. Voice assistants, automated workflows, and agent-to-agent communication all represent this trend toward invisible interaction.
Data Over Decoration
That stunning landing page animation you spent weeks perfecting? An AI agent will never see it. That carefully crafted visual hierarchy? Meaningless to an algorithm. What matters now is structured data, clear labels, and logical organization—the invisible scaffolding beneath the visual surface.Think of it this way: you're no longer designing a magazine layout; you're designing a filing system. Every piece of information needs to be clearly labelled, consistently formatted, and logically organized. An AI doesn't appreciate beautiful design, but it definitely notices when your form field is named "input_37" instead of "customer_email."
Speed and Predictability
Human users might tolerate a loading spinner or forgive a slightly inconsistent interface. AI agents won't. They operate at machine speed and have no capacity for patience or intuition. If your API takes three seconds to respond, a bot will timeout or bombard your server with retries. If your button IDs change randomly between page loads, automation scripts will break.One colleague learned this lesson the hard way while trying to automate a simple data download from a government website. The interface was unpredictable—buttons shifted position with each search result, fields lacked clear labels, and the workflow kept changing. What should have been a five-minute automation task became days of frustration as scripts kept clicking the wrong elements and entering data in the wrong fields.
The Invisible Interface
Perhaps the most profound shift is that much of the "interface" is becoming invisible. APIs, structured data formats, and machine-readable documentation are the new UI elements. Your website's HTML structure matters more than its CSS. The way your endpoints are named and organized becomes as important as your visual navigation menu.This is sometimes called "Zero UI"—the idea that the best interface might be no visible interface at all. While the term itself has become controversial since being coined in 2015, the underlying technologies have gone mainstream. Voice assistants, automated workflows, and agent-to-agent communication all represent this trend toward invisible interaction.
The Dual Reality: Designing for Humans AND Machines
Here's the good news: we're not abandoning human-centered design. The future isn't an either-or choice between human UX and agent AX—it's both simultaneously.
This dual reality creates fascinating design challenges. Consider a medical application: doctors need to see visual charts and simplified summaries for quick decision-making. But the AI agent accessing the same data needs structured JSON, detailed metadata, and machine-readable relationships between data points. One application, two completely different "user" experiences.
The emergence of the Model Context Protocol (MCP) in late 2024 showed how seriously the industry is taking this challenge. Created by Anthropic as an open standard for connecting AI agents to data sources, MCP was quickly adopted by OpenAI, Microsoft, and dozens of other companies in early 2025. Think of it as USB-C for AI connectivity—a standardized way for agents to access your services.
Designing for both audiences means thinking about accessibility in new ways. Those ARIA labels you added for screen readers? They also help AI agents understand your interface structure. That clean heading hierarchy you implemented for human readability? It helps algorithms parse your content efficiently. The semantic HTML you use for better SEO? It makes your site more navigable for bots.
Ironically, by designing for machines, we often end up with interfaces that are better for humans too. Clean structure benefits everyone. Clear labeling helps all users. Consistent patterns work universally.
Practical Guidelines for the Bot-Friendly Future
If you're building digital products today, here's what actually matters for AI accessibility:
Maintain Predictable Layouts Keep buttons, fields, and navigation elements in consistent positions. Avoid dynamic layouts where elements jump around based on context. A predictable interface helps both automation scripts and users with cognitive disabilities.
Use Clear, Descriptive Labels Every interactive element needs an unambiguous identifier. Instead of generic IDs like "button_7" or "input_field," use descriptive names like "submit_order_button" or "billing_address_input." Both humans reading your code and AI agents parsing your DOM will thank you.
Provide Machine-Readable Structure Use semantic HTML elements properly—header, nav, main, article, aside, footer. Structure data in clean tables with proper markup. When possible, offer data export options in formats like CSV or JSON rather than forcing users (human or AI) to scrape displayed information.
Optimise for Speed AI agents don't watch loading spinners. They timeout. Ensure your pages and API endpoints respond quickly. Aim for sub-second response times wherever possible.
Test with Automation Tools Don't assume your design works for bots—actually test it. Run Selenium scripts or use RPA tools like UiPath during development. If your automation fails to find elements or complete workflows, fix those issues before launch.
Make Errors Transparent When something goes wrong, provide clear error messages that both humans and machines can understand. Don't just highlight a field in red—include text that explains "Password must be at least 8 characters" or provide an error code that automation can parse and handle appropriately.
The Bottom Line
The future of UI design isn't about abandoning visual interfaces—it's about expanding our definition of what an interface can be. Sometimes it's a beautifully crafted screen. Sometimes it's a well-documented API. Sometimes it's structured data embedded in HTML. Often, it's all three simultaneously.
The principles that made you a good designer still apply: clarity, consistency, efficiency, and empathy. You're still solving problems and removing friction. You're still making complex tasks feel simple.
The difference is that now you're solving these problems for two very different types of users: the humans who set goals and make decisions, and the AI agents that execute tasks on their behalf.
Your website might get zero clicks from humans and still be a massive success because the agents love it. Your carefully crafted visual design still matters for trust, branding, and those moments when humans do check in. But underneath that visual layer needs to be solid, structured, machine-readable design.
Welcome to the age of the invisible interface. It's not replacing traditional design—it's adding a new dimension to it. And that makes our work more complex, more challenging, and ultimately more important than ever.
After all, when your biggest user has no eyes, you learn to design in a whole new light—or perhaps more accurately, in a productive and fascinating darkness.
