Designing for Two Audiences on a Two-Track Internet

Insights

Stay in the know

Last week, I came across a homepage for a video production company. It scrolled for what felt like a mile. Every service, every keyword, every question a buyer might ask, stacked one on top of the other. It was clearly written for AI agents, and as a human reading it, I honestly didn't feel compelled to hire them.

I don't share that to call anyone out. I'm seeing a lot of sites like it right now, and I understand why they exist. But I kept thinking about that website, and it eventually set up a conversation I had this week with Senior Designer Brittany Alford on our LinkedIn Live about what she calls the two-track internet.

A Website With Two Audiences

For most of the time we've been in business, a company's website was its most important piece of owned media. It was where your story lived, where customers came to learn about you, and, if you did the work well, where they decided to choose you. A lot of what we built for clients was about how a site felt, because the first impression mattered.

That's no longer the whole picture.

More people now get their answers from ChatGPT, Claude, or Perplexity without ever visiting the sites those answers come from. Bots already outnumber people online; Imperva's latest Bad Bot Report puts automated traffic at more than half of all web traffic. Every organization we talk to is working out how to show up in AI results, and so are we.

Brittany brought me the idea that framed our conversation. She'd read a Digiday piece about an experiment at The Economist exploring the idea of a two-track internet, with one experience designed for people and another for AI agents. Rather than asking a single page to serve both, they're asking how to serve each well on its own terms.

In her view, that doesn't have to mean two websites. "Ideally it would be one place to manage content that just gets interpreted in different ways," she said. A browser asks for a web page. An agent might ask for structured data. Same source, different outputs.

The distinction matters because the two audiences want very different things. People respond to story, emotion, clear hierarchy, and brevity. Agents want density, context, and explicit facts. A clever headline that works on a person is just ambiguity to a machine. Try to satisfy both on one page and you get that mile-long homepage, which serves neither audience particularly well.

Found, Trusted, Chosen

One viewer asked whether designing for agents goes beyond adding an llms.txt file. It does. That file is a reasonable first step, but it's only a first step.

After a couple of years of answer engine optimization work for clients, here's how I'd describe what we're seeing. I'll caveat that we're learning alongside everyone else. People and machines use your web presence in three ways: to find you, to trust you, and to choose you.

Getting found is the familiar part. Can an AI read your content and understand it in context?

Trust is turning out to be the part that matters most. AI reads what you've published, then goes looking for off-site signals – press, reviews, mentions, links – to decide whether you're worth recommending. One side effect surprised us. We're seeing companies get cited for blog posts they wrote seven years ago that no longer reflect who they are. The AI found them, treated them as signals, and surfaced them anyway. Managing your reputation online used to be a nice-to-have. It's becoming core work.

Consider what happens when your AC dies in August. You skim a few reviews, text a friend, and call whoever can come soonest. An AI asked to do the same research reads everything, including sources you've long forgotten.

There's a cautionary tale here, too. A few weeks ago, Nick and I were researching agencies in a particular discipline, and one firm kept getting recommended. We'd never heard of them. It turned out they had built several "best of" list sites ranking themselves first, scoring their competitors in the 30s and themselves in the high 90s. The AI took it at face value. I expect the models will get better at spotting that kind of gamesmanship. Trust you have to defend later isn't trust worth building.

Being chosen is the part that's least clear to me. Agents are starting to complete transactions on people's behalf, and ecommerce is already there. When agents are routinely filling out forms and booking calls, the paths they move through will become a real part of the funnel. We're still early in understanding what that means.

Websites Aren't Going Away

I get asked fairly often whether websites will exist in five or ten years. Brittany's answer is the one I keep coming back to. "I don't think websites will disappear," she said. "The customer will just land on them later in the journey."

We've seen versions of this before. Social media and Google Shopping didn't replace websites; they changed when people arrived. AI looks similar. Someone asks for three options that fit their needs, then visits those three sites to get a sense of each brand. Do their values align? What actually makes them different? Do they seem trustworthy?

I asked Brittany how she'd approach a site she knew only five people a month would see. Her answer was characteristically practical: it depends on what those five people are doing. If they arrive after an agent has done the research, they may be the most qualified visitors you've ever had. That's a different design problem than trying to impress a thousand casual browsers, and arguably a better one.

Brand Matters More, Not Less

Over years of branding work, I've noticed that many organizations don't treat their brand as seriously as they could. They refresh it every five or ten years, then start bending the rules almost immediately because everyone is moving fast. Generative tools have made that easier to do.

I think the agentic web is going to force the issue. If an AI is going to recommend you, it has to understand what makes you different, who you serve, and the value you bring. Polish and storytelling used to cover for positioning that wasn't quite clear. That cover is getting thinner, because the agent isn't watching your sizzle reel. Organizations with disciplined brands are likely to see outsized results. When your brand lacks clarity, AI won't create it for you. It'll amplify the confusion.

Brittany took this a step further into design. When an agent is the one describing you, you give up some control over how you're represented. So you need to give it the right materials: a clearly defined voice and tone, and visual components it can use accurately, whether that's an image in a chat window or a product card.

The Human Track as a Brand Activation

The part of this I'm most optimistic about is what happens to the human side of the web.

In marketing, a brand activation usually means a live event – a presence at a trade show or a stadium. It's a short, memorable moment designed to leave an impression. I can imagine that becoming normal online. Your core web infrastructure handles the agent track: structured, technical, and continuously maintained. The human track becomes a series of experiences built for specific moments, such as a new service line, a seasonal campaign, or a story worth spending time with.

A longtime viewer put it well in the chat: what if brand websites become the activation, focused on storytelling, emotion, and visual interaction? If agents handle the functional research, the website is free to do what people have always wanted from it.

Where This Leaves Us

We closed by imagining 2035: a web full of companies with flawless AI presence and no human proof. No faces, no stories, no community. Agents stop recommending them because they all sound the same, and the brands that win are the ones that are unmistakably human.

Brittany thinks we're already seeing early signs. People recognize the voice of an LLM, and it erodes trust quickly. Even design trends are bending back toward visible human touches. I'm skeptical of the more extreme predictions about AI replacing websites entirely, but I'm confident about this part. I can't picture a world where people stop wanting to connect with other people.

Nobody has this figured out yet, including us. What we can do is stay curious, keep experimenting, and build with both audiences in mind on purpose. That means giving agents complete, well-structured information, giving people something worth their time, and doing the slower work of earning trust in both places.

That's always been the job. The two tracks just make it easier to see.

Thanks to everyone who joined us live, and to Brittany for bringing such a good topic to the table. The full conversation is above, and you can catch the next one live on LinkedIn. If you're working through any of this, we'd be glad to compare notes.

Celebrating 10 years of Uncommon Thinking

Morning Coffee

Webflow Conf 2026 recap: what Source by Webflow means for your brand and website

Insights

Related Engagements

Marketing Website

A clean, credible Webflow site. Built fast, with purpose.

Brand Identity Refresh

A fresh look for your organization in just 6 weeks.