Matt Barbelli.
People aren't trawling through endless Google results anymore, they're asking ChatGPT for recommendations, getting AI summaries, shortlists and even direct answers. By the time they land on a website, they’re not browsing, but ready to act. But just because we can personalise the web experience from the beginning, it doesn’t always mean we should.
Think about it like walking into a restaurant for the first time and being greeted with: "Hi Matt, I know you like Shiraz and I know you like your steak medium-rare." It's specific, it's accurate, and it's absolutely unsettling. Why? Because it skips the natural rhythm of relationship-building and jumps ahead to familiarity that hasn’t been earned.
But that's what’s happening online with greater frequency. We're skipping the bit where relationships actually develop from a natural cadence, and jumping straight to the overly familiar uncle at Christmas who knows too much about your personal life.
Now imagine going back to that same restaurant for a second visit, and the waiter says: “Welcome back, Matt. Would you like the Shiraz again tonight?" That feels completely different. It's warm, human, and shows someone's paying attention without pretending to know everything about you. It's the difference between surveillance and service, creepy and caring.
In 1970, Japanese roboticist Masahiro Mori proposed the uncanny valley hypothesis which suggests that as robots or computer-generated characters become more human-like, our emotional response initially becomes more positive, but at a certain point, the response turns negative, creating a sense of unease.
We’re hitting that same middle ground with AI on the web, they're personal enough to feel invasive but not authentic enough to feel trusted. But the reason for investing in a good customer experience has always been about building trust and the same principle should guide how we use AI on the web.
Greeting users with ‘Welcome back, CFO of XYZ Corp. Based on your browsing history, here’s a case study you haven’t finished yet!’ - might be technically impressive but it often feels wrong. It’s not that users don’t want relevance, they just don’t want to feel watched or worse, profiled.
Agencies are part of the problem here. Every agency says it, every pitch deck promises it - ‘We put humans at the centre’. But if you're showing first-time visitors hyper-personalised content based on data, you're not putting humans at the centre, you're putting your algorithm there.
Real human-centred design embraces how people actually behave online. It allows for serendipity and lets people surprise themselves, not just confirm what an AI thinks it knows about them.
To truly design for humans we need to embrace ambiguity by allowing for discovery and creating experiences that feel natural and not robotic. And when we do use AI (and we should), we need to do it in ways that support the relationship not simulate one
The future of web experience isn't about building digital stalkers. It's about building digital companions, tools that get smarter as relationships develop, not systems showing off from day one.
I’m not saying that AI won’t make websites better, it absolutely will. But striving to deliver the best digital experiences for customers won’t come from the ones that know everything, but the ones that start listening. So let's embrace AI, personalise and anticipate, but let’s not get creepy with it.
Matt Barbelli, founder and managing director at Wonderful