What Happens When We Start Designing for AI Instead of People? | by Anna Li | Jun, 2026 | MediumSitemapOpen in appSign up<br>Sign in

Medium Logo

Get app<br>Write

Search

Sign up<br>Sign in

What Happens When We Start Designing for AI Instead of People?

Anna Li

5 min read·<br>8 hours ago

Listen

Share

Press enter or click to view image in full size

Photo by Joachim Schnürle on UnsplashWe adapted our content for search engines. Are we about to do the same for AI?<br>For years, we optimised websites, apps, and online content for search engines. Most of us accepted that trade-off.<br>A few extra keywords here.<br>A slightly awkward heading there.<br>Maybe a page written more for Google’s algorithm than for an actual person. It wasn’t always ideal, but it was understandable. Visibility mattered.<br>Now I can’t help wondering whether we’re about to do it again.<br>Only this time, the audience isn’t a search engine.<br>It’s AI.<br>And the question that keeps bothering me is surprisingly simple:<br>What happens when the things we build online are no longer designed primarily for humans?<br>Information Architecture Was Built Around Human Understanding<br>One of the ideas that has always fascinated me about information architecture is that it isn’t really about navigation menus.<br>Or site maps.<br>Or labels.<br>At its core, information architecture is about helping people find, understand, and use information.<br>Rosenfeld and Morville describe information architecture through concepts like organisation systems, labelling systems, navigation systems, and search systems. The goal isn’t simply to help users locate something. It’s to help them make sense of where they are and what they are looking at.<br>In other words:<br>Information architecture was built around human understanding.<br>Human mental models.<br>Human behaviour.<br>Human uncertainty.<br>Human mistakes.<br>When we evaluate an information architecture today, we ask questions like:<br>Can users find what they need?<br>Do the labels make sense?<br>Can people navigate confidently?<br>Does the structure match their expectations?<br>They’re all fundamentally human questions.<br>But what if our future users aren’t only human?<br>AI Doesn’t Need Information the Way Humans Do<br>Imagine you’re looking for a hotel.<br>You might compare locations, read reviews, look at photographs, check cancellation policies, open five tabs, and change your mind three times before making a decision.<br>A human journey is rarely linear. The process itself creates confidence.<br>An AI agent approaches the same task very differently. It doesn’t care about beautiful navigation. It doesn’t care about visual hierarchy.<br>It doesn’t care whether your menu structure feels intuitive.<br>It simply wants information.<br>Structured.<br>Consistent.<br>Accessible.<br>Machine-readable.<br>Done.<br>The things that make an experience great for humans aren’t necessarily the same things that make information easy for machines to consume.<br>And that’s where things start getting interesting.<br>We’ve Seen This Movie Before<br>This is the part that makes me slightly uncomfortable.<br>Because we’ve already watched something similar happen with SEO. Companies didn’t intentionally decide to make digital experiences worse. They simply followed incentives.<br>Visibility mattered.<br>Ranking mattered.<br>Traffic mattered.<br>So content gradually became optimised for search engines.<br>Sometimes that improved the user experience.<br>Sometimes it didn’t.<br>Content farms appeared. Keyword stuffing appeared. Pages were written to satisfy algorithms rather than readers. Eventually search engines adapted.<br>But it took years.<br>Now we’re entering a world where visibility may increasingly depend on AI systems.<br>And history suggests that organisations will follow the incentives.<br>They always do.<br>The Metric Problem Nobody Is Talking About<br>Imagine a company a few years from now.<br>Their analytics look fantastic.<br>More traffic. More referrals. More product recommendations. More discoverability. Everyone celebrates.<br>But then someone asks an uncomfortable question.<br>Who generated those interactions?<br>Humans? Or agents?<br>If AI agents become better at navigating systems than humans, our metrics may start telling a very incomplete story.<br>The numbers could improve while the human experience gets worse.<br>And we might not even notice.<br>Because historically, success metrics have assumed that a user is a person.<br>What happens when that’s no longer true?<br>Could We Accidentally Optimise Against Humans?<br>This is where the issue stops being technical and starts becoming ethical.<br>Let’s imagine that AI agents become a major source of traffic, recommendations, purchases, and customer acquisition.<br>What happens next?<br>Companies will invest where they see returns. That’s rational. That’s expected.<br>But if designing for AI becomes more profitable than designing for people, where does the investment go?<br>What gets prioritised?<br>What gets neglected?<br>What happens when making information easier for agents makes it harder for humans?<br>What happens when the easiest path to visibility isn’t the best path to...