🎙️AI Changed The Web. How To Build For It?

Hi everyone – hope the weekend's treating you well. It’s interview time! Subscribe to our YouTube channel, or listen the interviews as podcasts on Spotify or Apple.

At some point in the last year, bots became your biggest website visitors. Not people. Not crawlers. Not even APIs. Bots with goals. Agents with plans and their own agenda.

Linda Tong, CEO of Webflow, has seen it up close – and she's redesigning the web to meet them. In this episode, we talk about what it means to build agent-first websites: How to talk to bots. How to let them click buttons. And how to create experiences that work for humans and AI – without turning the internet into garbage.

We cover:

• When bot traffic started overtaking humans

• Why AEO (agentic engine optimization) is the new SEO

• Why websites need a second language – for LLMs

• What "agent-ready" structure really means

• Hybrid UX: visual for humans, semantic for agents

• Why dynamic, personalized web experiences are overdue

• Leadership, kindness, and Ender’s Game as a design philosophy

This one's fast, nerdy, real, and fun. Linda’s not afraid to challenge old assumptions – or to break her own product if it means building what’s next. Watch it now →

The transcript (edited for clarity, brevity, and sanity. Always better to watch the full video though) ⬇️

Ksenia Se: Hi Linda, and welcome to Inference by Turing Post. I’m very happy to have you here.

Linda Tong:
Very stoked to be here. Thanks for making time – I’m excited to chat with you.

Ksenia: You're leading Webflow, a $4 billion no-code web development platform, through an amazing moment. It's a full-on AI transformation. So here’s the question: there’s a new species online now, right? Not quite APIs, not quite people – more like, as Karpathy called them (and I love it), “human spirits on the internet.”

So is it true that traffic online isn’t just humans anymore? Do you see these spirits? And when did this shift begin?

Linda: Absolutely. What people are often looking for is a hard number – like, what percentage of traffic is human vs. bots or spirits, whatever you want to call them. But it really varies.

Some sites are still 90% human, 10% bot. Others? The reverse – 90% bot, 10% human. So it’s not consistent, but I am seeing a clear trend: bots or chat agents are increasingly responsible for a big chunk of site traffic.

And not just passive traffic – they’re showing up in the top-of-funnel referrals. We’re starting to see more and more of that traffic being driven by bots that actually direct people (or other bots) to different parts of the internet.

Ksenia: When did that start?

Linda:
Over a year ago. We started seeing early signs at the beginning of last year – little spikes here and there. But by last summer, it became disruptive.

And since then, it’s only grown. If I follow the trend lines, I think bots and agents will make up the majority of internet traffic by the end of this year. So it’s not exactly new – it’s been quietly building.

Ksenia: Interesting. So what do you do with that? Does it change your strategy in a meaningful way?

Linda:
Absolutely. First, it’s really important to understand what this is – what bot traffic actually means, and how we should think about it.

There are good bots and bad bots. We've always had bots – that’s not new. Bad bots do nefarious things like DDoS your site. But now, we have good bots too. These can come from major LLMs, or from agents trying to index your site, understand it, and reference it in response to user queries in chat.

So you want them to be able to crawl and understand your site. And as more companies build their own models and their own chat-based search tools, we’re seeing more of these agents crawling the web. Overall bot traffic – good and bad – is increasing.

People ask me, “Should I just block it?” And I say: no, don’t block it. That’s like blocking Google back when SEO was all about optimizing for search. What we’re seeing now is the evolution of SEO into what people are calling AEO, GEO, or AIO – ways to optimize for AI-driven experiences and generative engines.

Being indexable, referenceable, understandable to these bots is key. You want them to be able to read your site – not just access it, but actually parse and represent it well.

And just like we used to reverse engineer Google search to climb the rankings, we now have a new challenge: reverse engineering how these chatbots and agents rank or reference content. And it’s a different system. Things like site structure, content distribution, content freshness, and external references (like Reddit or social platforms) all matter. So optimizing for this new kind of visibility means adapting how we structure and publish online.

Ksenia: Right. It’s still so early – we’re figuring out how to interact with these agents and how they can interact with software. Do you think something like llmps.txt, instead of raw HTML, CSS, JavaScript, could be a way to speak to them directly? Also, agents don’t use clickable URLs, right? That has to change.

When you think about infrastructure shifts – and your job, which is helping others build websites – what’s your action plan for building websites for agents?

Linda:
It’s early days, but I really believe that just like mobile changed everything around 2009–2010, we’re now entering the era of agent-optimized websites.

Back then, everyone realized they needed mobile-optimized sites. Now, we’ll need agent-optimized sites – sites that are built specifically to interact with LLMs and agentic workflows.

That means rethinking the structure entirely. These agents don’t interact the way humans do – but they’re still trying to accomplish similar goals. Today, websites are built for human experiences. But we need to start thinking of agents as a new kind of audience, with their own interaction patterns.

At Webflow, we’re thinking hard about what an agent-first site looks like. How do you make it seamless? Because building for humans is already hard – mobile, tablet, desktop, different screen sizes, content layouts. Now we’re adding another layer.

And no, we can’t just skip human experiences and build only for agents. So it’s about making this complexity manageable and building in a way that serves both.

Ksenia: So what’s most important when building for this new audience? What do agents need?

Linda:
The biggest thing is making it easy to reference data.

Sites with deep, tangled architecture, mixed publishing models, or scattered content – those are hard for agents to parse. These bots are trying to understand your structure, your brand principles, your navigation, your content.

So we need better ways to surface that information – like site maps, but more dynamic and structured for agents. Right now, a lot of human-facing design is about delight, storytelling, rich visuals – and that stuff doesn’t translate.

When agents scrape a site, they might hit every page but skip the actual meaning. That’s wasteful and inefficient. We need to offer a simplified, structured view of what’s on a site – and how it’s connected.

You also mentioned links, and that’s a big one. Your site is just one part of your online presence. Agents are trying to understand your whole brand, not just one domain. So you have to represent where else you live – your social profiles, media, partner content, etc. Stitching all of that together gives agents a fuller picture.

And third, freshness matters. Recency is heavily weighted. Sites that are updated regularly with real, opinionated, human-written content rank better than stale ones.

So brands need to move faster – not just to stay relevant, but to be understood. Relevance and dynamism go hand in hand in the world of AI agents.

Ksenia: That’s all about reading and understanding. What about action? How do you make the agent actually click a button?

Linda:
Yeah. I mean, part of it is user experience. It's a mix of: one, can you simplify your user experience? Whether it’s APIs or building an MCP that agents can interact with to take specific actions on your site or product. Right now, people are using MCPs mostly to build capabilities for agents to interact with their site.

But it’s still convoluted. The MCP is usually layered on top of the human user journey. APIs were really designed to go directly to the source – call the action you want, interact the way you want. So, I think there’s a real need to understand which APIs you want to expose to an agent, and maybe publish a sort of manifest – here are all the actions you want to make available, and here are the guardrails around them.

That’s what people are trying to simplify with MCPs. It’s probably a midterm solution. Longer term, I think agentic platforms that are hooked directly into those actions are what’s going to take hold.

Ksenia: Do you think we'll end up with dual outputs – visual and user-facing for humans, semantic and structured for LLMs and agents?

Linda:
I actually think it’ll be more of a hybrid. The reason AI has taken off the way it has, the reason it's disrupting how we work, is because the human-specific interfaces just weren’t enough. If they were, we wouldn't be seeking alternatives.

People complain about enterprise SaaS software all the time. You just want to update one thing or file an expense report – really simple stuff. And you’re like, “Why is this so hard?” And now with AI, you’re like, “Oh, this could be so much easier.”

So yes, there's a real need to simplify those workflows. But I don’t think we’re at a place yet where people fully trust an agent to just go and do everything for them. There’s still human-in-the-loop interaction. I think we’ll be in a hybrid world for a while.

The dream, of course, is full automation – agents doing everything while we go watch Netflix. But in reality, we’re going to want agentic workflows, paired with human-facing experiences that are visual and easy to understand. That way, collaboration with agents still feels natural and intentional. The agent handles the overhead – the tasks – while humans step in where it really matters.

Ksenia: That’s exactly what I was thinking before our conversation. I was playing with a clog, trying to build a website. It’s not perfect, but you can tell it, “No, I don’t like this color. Give me the color of the sea in the morning sunlight.” And it gives you something pretty close.

So on Webflow, you’ve got the AI site builder. And it has a prompt – you can edit it – but it’s not really a conversation. Do you think Webflow could become conversational? Merging agentic, human, and dialog-driven interaction?

Linda:
I can’t announce anything, but… good question.

Our vision as a company – just to rewind – Webflow’s been around since 2012. The vision is actually pretty revolutionary: we want to bring developer superpowers to everyone.

We get branded as a no-code platform, because that was the movement at the time. But the real thesis is that everyone should be able to build for the web. Everyone should be able to create. Human creativity – that’s the power we’re trying to unlock.

We started with a visual builder so people who didn’t code could build powerful websites. Now with AI, we’re accelerating that vision. Instead of just visual tools – or maybe paired with them – you can now articulate what you want. You can speak it. You can type it.

And I think we’ll go further. I might say: “Make this site feel like a sunset in Athens, Greece – something Mediterranean.” And it’ll get me something close. But I might also want to interact with it in new ways.

So far, LLMs let us type or speak – but what if I could draw something? What if I could visually interact with it to spark creativity? I don’t know what that future format will be, but I do believe AI should superpower what you can create with Webflow.

Visual is one mode. Natural language is another. And I think the hybrid of the two could be really powerful. We're working on a lot of that.

Ksenia: Can you paint the big picture for me – maybe in a few years, maybe in just a few months – what does building a website with AI look like in your ideal world?

Linda:
Building a website with AI is already here, and it's only going to get better. But I actually think that’s not where the real challenge lies. We’re in this creation phase right now – with AI, you can build a site, an app, a picture, anything. And ideally, it should feel like a conversation.

You might say, “I’m building a business. I want a website that reflects the power of my brand, evokes this kind of emotion, looks like this, and includes these elements.” We can already do that today – the results vary, of course. Sometimes it’s the prompt, sometimes it’s the tech. But we’re getting close to the point where what you articulate and what you get are almost perfectly aligned.

Where we’re not spending enough time is asking: once you’ve built the website, then what? And this is where it gets exciting. The web is changing – not just because of the last two or three years, but because we’ve been on a transformation for a while.

We used to publish static websites where millions of people had the same experience. Which is kind of absurd, right? Think about e-commerce. If you walk into a physical Gap store, it’s laid out the same for everyone – same racks, same items, same layout. That makes sense because it’s a physical space.

But online? There are no physical constraints. There’s no reason we should all see the same version of a website. The experience should be dynamic. And people have asked, “Why can’t I make it dynamic?” But the answer has usually been: “I just don’t have the time.”

Ksenia: That’s true.

Linda:
It takes so much just to get the website built and live. Operating it is already a lift. So the next step – building a dynamic experience – often feels out of reach.

Now, with AI, that’s changing. We can move faster. We can make the web dynamic. That means you and I could have completely different experiences on the same site – based on location, whether you’ve visited before, or what I know about you.

Even better, the site can evolve every time you come back. That’s what excites me most – not just building a site, but having that site transform to fit its audience. Hyper-personalized experiences, tailored to each visitor.

Because every visitor is different. They have different goals. And if I’m building a site for real interaction, I want to meet those goals in a personalized way.

Ksenia: Yes – and when you layer on wearables or glasses, the whole visual experience changes again. That’s such a different interface and view. It’s exciting.

Linda:
Exactly. This is why I think the next leap is going to be about dynamic experiences. Right now, we’re all obsessed with what we can create – but once we get over that hump, we’ll be able to focus on how it adapts.

I’ve built websites on every service. I’ve made 70+ apps using different app builders. But the truth is – most of them are just internet garbage. I made them, and they’re just sitting there.

Only a small subset are actually used. But those are the ones I want to nurture. I want to build them out, make them sustainable, turn them into real products.

This next phase is the most exciting – where we go from creation to value. At Webflow, we want to help people build sites, yes – but more than that, we want to make it easy for those sites to deliver long-term value.

That means thinking about the whole lifecycle:
– How do you create dynamic, personalized experiences?
– How do you leverage insights?
– How do you run experiments?
– How do you manage your content and assets?

That’s the cool challenge.

Ksenia: That’s super, super exciting. I wanted to go back to something you mentioned earlier – SEO. Everyone’s still optimizing for SEO. But is SEO dead? What comes next? And how do you explain to people that maybe they should be optimizing for something else?

Linda:
It’s funny – maybe this is aging me – but no, I don’t think SEO is dead. It’s just convenient to declare things dead every time a new technology comes along.

I remember when Spotify came out, everyone said, “Radio is dead.” But radio is still a multi-billion-dollar industry. When digital streaming came out, people said, “Broadcast is dead.” It’s going stronger than ever.

So I don’t believe that one thing replaces another entirely. What we’re seeing is what always happens with technology shifts: we get a new way of doing something we’ve always done – in this case, search.

Now we’re in a world where you still need to think about SEO – that doesn’t go away – but you also need to carve out part of your brain for AEO or GEO. Right now, that means optimizing for chatbots. But whatever new format emerges next, you’ll need to adapt to that too.

All it does is add more work. It’s no longer a one-dimensional game. It used to be: just optimize your website for SEO. Now it’s SEO and AEO. And the big question is: does what you do for AEO hurt your SEO score? That’s what we’re thinking about a lot, and helping our customers navigate.

SEO has always been about things like keywords, backlinks, domain authority – how your site is referenced.

AEO is different. It’s a lot more weighted toward authenticity. It looks at your content’s quality, how authentic it is, how you're referenced across the web. Places like Reddit or Quora can boost your rating. It values freshness – how recent your content is, how often it’s updated. And it looks at how easy your site structure is to parse.

So you’re optimizing for two different sets of signals. Thankfully, we haven’t seen one actively harm the other – yet. But they are clearly diverging in how they define relevance and trust.

Ksenia: Sounds like we’ll need to create a guide. Do you have one?

Linda:
We don’t have an official guide – yet. But we’ve shared some best practices. We’ve run a couple of webinars and posted follow-up blog posts with key takeaways.

The reason we haven’t formalized a guide is: unlike most tech changes, this space is shifting every few days. New models keep coming out, and each one thinks differently.

AEO isn’t consistent. You have to consider how Gemini handles it versus how ChatGPT does, versus Claude. Each one has a different methodology, different weightings.

So we’re still in early days as a society in understanding how this all works. Even the models themselves are continuously adjusting.

The people who will win here are the ones who keep testing, keep learning, and iterate fast. We’ve got a few best practices – but who knows if they’ll still be best practices three weeks from now.

Ksenia: We don’t have much time, so I want to switch gears. We’ve been talking a lot about agents, about agents talking to agents, and even AGI and superintelligence. What is AGI to you? And if it happens, will it be able to design, code, deploy products and websites? Would it use something like Webflow – or replace it?

Linda:
I’ll zoom out a bit – not just thinking about software, but about the role of AI in human life.

My personal belief is that AGI will be an accelerant and a partner to humans. What makes us unique as a species is that we’re creative, opinionated, and constantly changing. AGI will reflect human intelligence and synthesize it at scales we’ve never seen. It’ll be able to iterate and create – it might mimic some human behaviors – but I don’t believe it will replace us.

Now, applying that back to software: AGI will absolutely change how we interact with tools to accomplish tasks. It might be able to do most of the work. But I still believe it will need some direction – some human perspective.

Will Webflow exist in that world? I hope so.

Will it exist in the form it does today? Probably not. It will be fundamentally different. But the mission – to help people tell stories, build digital experiences, and create – that has lasting value.

Ksenia: It’s hard to stay flexible in times like these.

Linda:
It is. You have to be fearless.

What does that mean for how we operate as a business? It means we’ll have to disrupt ourselves. And that’s painful. It’s scary. There are risks – revenue, relevance, everything. But I believe that if we stay focused on doing something valuable for people and for businesses, we’ll find a way through it.

It’ll just be a very interesting journey.

Ksenia: Let’s stay human for a moment. Can you tell me what book shaped you most – in business or life?

Linda:
You’re going to think I’m an alien now. My favorite book – and series – is Ender’s Game by Orson Scott Card. I reread the whole series every year. I’ve been doing it since elementary or middle school.

The core series is Ender’s Game, Speaker for the Dead, Xenocide, and Children of the Mind. Then there’s a parallel series called Ender’s Shadow – about eight books total, not counting some of the more recent additions.

Ender’s Game is about a six-year-old boy who’s trained in space to save humanity. He and other brilliant children are taken to a place called Battle School, where they’re trained in war, theory, and leadership to fight an alien species called the Buggers.

What’s so compelling is that you see leadership development through the eyes of a child. Humanity decides that children – not adults – are best suited to protect the world. They’re manipulated, yes, but also taught strategy and theory.

Ender, despite being the youngest and smallest, becomes the leader everyone follows. He’s brilliant. He’s empathetic. And he takes on this enormous burden of responsibility. One of the most beautiful lines in the book is:

“He could never defeat an enemy unless he loved them – because only when you love someone do you truly understand their weakness.”

That always stuck with me.

The book is fun and intense, but it’s also a profound exploration of leadership. It shows that true leadership isn’t about authority – it’s about influence. And Ender leads through kindness and understanding, not power. That’s what shaped me most.

Ksenia: You mentioned kindness. I’ve been thinking a lot about kind AI. That’s something we don’t talk about enough.

Linda:
Absolutely. Maybe that’s what we need to build next.

Ksenia: That would be wonderful. Thank you so much for this conversation – it was fascinating and a lot of fun.