Romain Colomer, Experience Director at CreateRecently during some R&D, I watched as an AI agent frantically searched a luxury retail website and then abandoned it without finding the product it was looking for. It happened quickly, but in that moment, I realised that as an industry — and more broadly as a species — we had crossed a profound threshold.
The agent wasn’t only searching — it was judging, deciding, and ultimately dismissing a digital experience that failed to meet its standards. For me, this signaled a turning point in how we might need to now approach digital interactions in an increasingly AI-centred world.
You might be thinking that this is purely anecdotal, yet the numbers behind this example are staggering. In 2024 alone, some of our clients saw AI-driven search traffic increase by more than 5000%, largely due to ChatGPT’s integration of search, which launched as recently as October.
While only a small percentage of these visits were likely agentic, they demonstrate the very real need to start thinking about designing platforms and content for AI optimisation. This is what some in the industry are terming Generative Engine Optimisation (GEO).
While still dwarfed by search engine traffic, AI sources are not to be discounted. Across our work, we’re seeing that AI-driven traffic results in visitors who are at least 20 per cent more likely to engage and spend at least 50% more time on site. Assuming that these numbers are still predominantly humans arriving via AI sources for the time being, we can conclude that as consumers, we are beginning to trust in AI’s filtering and decision making for us.
This shows that there is a kind of prequalification of choice going on. This is, of course, self-evident in content based recommendations on platforms like Instagram and TikTok, but this trend has also been growing for some time, with the search term “best” doubling back in 2017, indicating an ongoing default to AI systems knowing what is best for us.
But beneath these numbers lies a deeper transformation in how we interact with digital platforms. As such, we’re seeing the emergence of a need to start thinking with a dual-audience mentality, where digital experiences must simultaneously serve both human users seeking emotional resonance and AI agents (or summarisation algorithms) demanding hyper-efficiency and structured data.
Now, with OpenAI’s recent release of its first agent, ‘Operator’, we expect these numbers to increase exponentially, especially as global rollout begins and user adoption of AI continues to rise. Operator essentially takes all the benefits of ChatGPT and turns them into self-reliant, actionable systems that make decisions and solve problems (for the most part) without human intervention, meaning users can now do everything from weekly grocery shopping to booking flights, and even picking stock and crypto trades (although OpenAI has some basic guardrails around this). Nonetheless, this ease of use and streamlining of products and services will have a big impact on how we all interact with the internet and digital experiences as a whole.
As Professor Ethan Mollick of The Wharton School recently put it, “One thing academics should take away … is that a substantial number of your readers in the future will likely be AI agents.”
This duality creates interesting paradoxes in digital design. Recently, when testing Operator, I prompted it to navigate a well-known, award-winning website renowned for its immersive user experience and bold brand storytelling.
Unfortunately, however, the agent didn’t wait for animations to load or appreciate the carefully crafted interface — after attempting to reload the site several times, it simply left to complete the goal elsewhere. This behavior brutally illustrated the growing tension between immersive digital experiences and the ruthless efficiency demanded by AI.
In a sense, however, this experiment highlighted that some common user challenges in digital experiences may now be amplified tenfold. Simply put, as much as we designers hate to admit it, complex interfaces, slow load times, and forced story-scrolling can often be frustrating for humans too. If anything, the shift toward AI centred design may even be a revival of human-centred design (HCD) and require designers to strictly adhere to core tenets of HCD in order to create platforms that continue to deliver impact.
Nonethless, I suspect that what will emerge from this is a new paradigm that challenges our basic assumptions about digital experience design. While we’ve spent years optimising for mobile-first experiences, we’re now entering an era that demands AI-first thinking.
Looking ahead, organisations will need to evolve to create platforms that can dynamically adapt to serve both audiences effectively, without compromising the experience for either. Not to mention beginning to plan for what an agentic-internet will do to online metrics like ROAS, abandoned-cart rates, traffic numbers, and so on. Organisations that fail to adapt risk falling behind, or at minimum becoming invisible not just to AI agents, but to the humans who increasingly rely on them.
Further, as AI agents become more sophisticated in their ability to act on our behalf, we have to reconsider fundamental questions about the nature of digital interactions. When AI agents make decisions about what products we see, what information we receive, and what opportunities we’re presented with, they won’t just be mediating our experiences—they will help shape our choices and, dare I say it, increasingly challenge notions of free will.
Finally, as we approach this transformation, we may need to remind ourselves that the goal isn’t to choose between human-centric or AI-centric design, but to create experiences that can at once be both engrossingly immersive and, when required, simply invisible—where everything just works. After that, the big challenge will be ensuring that these interactions, interfaces, and systems are serving us—rather than us serving them.
By Romain Colomer, Experience Director at Create
