Is human attention still something we want to CAPTCHA?

The future of the digital user experience

Afonso Cruz & Wouter van den Berg

You've likely narrowed your eyes at a CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) or clicked all images that include a traffic light. Why? To prove you are human. However, being a human online isn't that important anymore.

The new principles of design

Are we at the beginning of shifting from human-centered to agentic-centered design? Design either creates behavior or responds to it. Websites used to hate bots, now they might encourage them to show up. We used to Google things, now we ask a Large Language Model (LLM). Search Engine Optimization has given way to Generative Engine Optimization (packaging things nicely for ChatGPT and company). In the past, principles of persuasion and some structure got your website further.

Now, AI-generated responses prioritize content clarity, accuracy and, of course, structure even more. While the two principles differ, there is some surprising overlap. Designing an inclusive website for people is also good for bots. For example, making a site accessible for a blind person also makes it more accessible to AI agents. 

So, who is designing for whom? A real-world example

Adrian Holovaty, creator of SoundSlice - a tool that reads sheet music - noticed users suddenly started uploading guitar tablatures instead of sheet music. His software didn’t support reading tablatures, so why were they doing it? Turns out, ChatGPT had been telling users to use SoundSlice for this purpose. It wasn’t true, and it made Holovaty's product look bad. Rather than add disclaimers that tablatures aren't acceptable input, Holovaty built the feature. “I’m glad to help people,” he said, “but it feels strange to develop features based on misinformation.” Ironically, ChatGPT’s hallucination revealed a real demand.

AI agents for search. What else?

Designing for agents isn't just about attracting the right attention, it directly affects the bottom line. Our AI avatars will do more things on our behalf, such as booking a flight and accommodation. AI agents cut out the middleman. This makes them a new sales channel, a part of your omnichannel strategy you might not have thought about much yet. While most of us are used to chatbots, the chat paradigm is limiting. Underlying APIs can be more powerful for specific tasks, without the annoying follow-up questions. Think small language models and macros for example.

Strategic questions for design:

• How can your services and APIs be made discoverable and useful for AI agents?

• How do you prevent misinformation from shaping AI behavior?

• Can your content survive and thrive when machines are the first readers?

• How can you best create protocols to track autonomous agent activity in the same way we track human user activity, and should you adjust your definition of digital touchpoint metrics?

• How can you design systems that guide both human users and AI agents, perhaps by treating AI agents as digital collaborators or “employees” who need onboarding, training and clear responsibilities?

  
  

Humans aren't out of the picture, but we’re no longer the only audience. The best digital experiences will be designed for humans and AI agents, accessible to both. In the age of AI, the true test isn’t proving you’re human; it’s designing like you know agents are watching.