The Internet of Intelligent Things

What happens when agentic AIs visit websites more than humans do?

According to Malwarebytes, 50% of all internet traffic is automated. As AI adoption grows and more and more time-consuming digital tasks are offloaded to AI agents, that 50% will inevitably increase. That thought should spark fear (or excitement) in the hearts of those whose business model revolves around display ads and click funnels.

A new paradigm is emerging that is going to bury those who don't adapt. At least, that's my take. It's difficult for me to see it any other way, unless AI adoption suddenly goes into decline, which seems unlikely.

People Doing Less

I told a friend once that AIs will write code and software engineers will just orchestrate them like a conductor. This was after seeing GPT-2 for the first time, when it was still relatively new. In about half a decade since then we've gone from inconsistent and incoherent chatbots, to AI systems that can build entire applications, do deep research across the web to produce detailed and coherent reports, as well as offer non-technical support and advice on all manner of topics and situations.

The growing capabilities of AI have obviously not gone unnoticed. Consumer adoption has been unprecedented. ChatGPT launched in late November 2022. By the end of January 2023 it already had 50M users. It's currently at around 800M. Enterprises were slow off the mark but are rapidly trying to become "AI-first" as they battle to get ahead of—or, at the very least, keep up with—competitors.

First-Hand Experience

I see this growth when I look at the adoption in my own social and professional circles too. And my personal experience, of course, is a prime example. Since I joined ChatGPT back in 2022, I've interacted with AI on a daily basis. Now, I'm using multiple AI services daily. From coding and writing tests, to research, design and ideation, I let AI do the heavy lifting whenever possible, so that I can focus on the more fun and interesting stuff.

I'm seeing the same behaviour in others, and it's no wonder, because it makes sense. Why would you spend hours or days traversing the internet for information when AI can do it for you in a few minutes? Or trawl through documentation for some very specific piece of info that AI could probably find in seconds?

Another example, just to drive the point home: I came up with a novel design for a piece of jewellery and I wanted to find out how much it would cost to get it made. I put not one, but two AIs to work to go do the research, considering different metals and techniques to achieve the structure and mechanism, distance from my location, quoted lead times and, of course, prices.

I used Manus and ChatGPT's Deep Research. Both spent about 10-15 minutes visiting numerous sites, extracting and evaluating data, compiling it, and eventually provided me with all the details I needed. If I'd done the research myself, between work, projects and other commitments, it would have been a minimum of days, but most likely, weeks.

Inadequacy of the Website

While 15 minutes to do weeks' worth of research is a massive productivity boost, it could still be better. Our websites were built (often poorly) for humans to navigate, not machines. Sure, we can add schema markup and other metadata to make it more machine-friendly, but the AIs still have to:

• wait for a UI to render
• hunt through navigation menus
• deal with pages heavily dependent on visuals
• find data scattered across multiple pages
• deal with cookie banners, modals, and other intrusive UI elements

Dealing with only one of the above is the best-case scenario, but access could also be entirely blocked by CAPTCHA-like tests of humanity. The bottom line? Our websites are not optimised for AI.

I think there's some irony here that the best way to improve the human experience is to improve the AI experience. And this is why I think a fundamental shift is coming to the web. Two emergent characteristics will change the nature of websites to such an extent that I think we'll probably need to use a different term for these "neosites". The emerging characteristics? AI-Accessible, and AI-Enabled.

A Brief Evolution of the Web

• We started with static websites in Web 1.0.
• We made them dynamic and data-driven in Web 2.0.
• Smartphones came along and our attention turned to mobile-friendly and mobile-first.
• Alongside the focus on mobile, we saw the emergence of the Internet of Things, real-world devices connected to the internet.
• Finally, we had the rise (and stall) of Web 3.0, which sought to bring decentralisation, blockchain, and data ownership to the forefront.

Now Artificial Intelligence is poised to unlock characteristics that will transform the internet again, but more dramatically than ever.

AI-Accessible Websites

For proof of what the future holds, we can look to the partnership between OpenAI and Shopify. ChatGPT, which has had product recommendations for a short while now, will gain access to Shopify's new Catalog, giving it real-time product information from millions of Shopify merchants. This type of integration means you could discover and purchase products online without ever leaving chat, or even opening a browser.

While OpenAIs Operator could theoretically hunt for and purchase products on your behalf, it's an inferior experience, because Operator:

• requires a $200 per month subscription
• uses human-style browsing to navigate the web and take action
• still requires manual input for payment details, etc.

The UI-based data discovery is the big bottleneck here, which the Shopify integration obliterates. Because Shopify has made its data AI-Accessible, ChatGPT can query it directly, bypassing search engines, UIs and clunky navigation for quick data retrieval and display. Waiting time is slashed. The human experience immediately gets an upgrade.

This is AI-Accessible in a nutshell: the site provides AIs direct access to rich, structured data; the AIs decide how best to display it to their users.

AI-Enabled Websites

At Microsoft Build 2025, Microsoft announced NLWeb, an open source project that:

[...] aims to make it simple to create a rich, natural language interface for websites using the model of their choice and their own data.

I tried it, and integration was...not simple. Still, while NLWeb might not yet work as advertised, it's an indicator of their vision of the future. I think they see what I see, and that's intelligence baked directly into websites. NLWeb seems to be a tool on the road to making that a reality.

But what does AI-Enabled give us? It gives us websites that don't just provide data, they contribute to problem-solving. In a sense, ChatGPT was the first true AIE website. Our current and most common experience of AI-Enabled websites is general purpose AI chat services, like ChatGPT, Claude, Gemini, DeepSeek... Of course, AI is the core value proposition for these services.

We also have the more traditional "dumb" chatbots, which were previously more annoying than useful. A site with a dumb chatbot doesn't qualify as AI-Enabled. But we're at a point in the evolution of AI that intelligent chatbots could become integral to elevating passive websites to something far more valuable.

Let me give a simple example. After receiving the reports from Manus and ChatGPT, I was left with a couple of short lists of jewellers, including links to their sites. It would be amazing to be able to visit one of the sites and ask questions specific to the jeweller, including current lead times.

Adding this type of intelligence to the site would be beneficial to the jeweller too. The AI could handle all the repetitive queries, potentially qualify leads, and even begin the conversation and extract information about design preferences and materials.

Or consider a non-technical user who needs to buy a new laptop. Instead of scouring the internet trying to figure out what to buy, they could go to the online retail site of their choice and simply start with, "I need a laptop". The AI would take the conversation from there, asking them about the purpose, budget, operating system preference, then present them with the best options. Low stress, high efficiency.

Neosites

The point here is that the web will shift away from being navigation-driven (search -> click -> nav -> click, click, click). The web will be intent-driven, where the human expresses their intent and AIs act to satisfy the intent. The companies who succeed in the "neoweb" (continuing the "neo" theme) will build intelligent sites that can respond to intent.

This becomes even more powerful when sites are both AIA and AIE. Theoretically, sites could make the "intelligence" accessible through an AIA interface or endpoint. This would allow AIs to communicate directly and work with each other to solve the user's problem.

At Google IO 2025, Google announced the A2A (Agent-to-Agent) protocol. It's designed to allow:

[...] gen AI agents, built on diverse frameworks by different companies running on separate servers, to communicate and collaborate effectively...

The infrastructure to allow "intelligent things" to communicate and collaborate across the internet is already being built. This is not theoretical or speculative. It's already happening. If I wasn't sure before, the recent moves by Microsoft, Google and other players give me more confidence than ever.

A new web is coming, and it will be defined by these two characteristics:

AI-Accessible: "Here's my product data, you figure out what to do with it"

AI-Enabled: "Tell me what you're trying to accomplish, and I'll help you solve it"

The Fallout

All of the above leads to the overarching question: If an "internet of intelligent things" is the future, what impact will this have on the web as we know it? I have some ideas, but none of them will be surprising given the progression I've already laid out. For this reason, I'll keep it brief.

What Changes

Display advertising becomes obsolete, in most contexts. In a world where AI agents browse the web more than humans do, the ROI on display advertising plummets.

Click funnels collapse, for the same reason as display ads. The intent-driven web brings the data to the human. What might be producing sizeable revenue now, will generate only a trickle.

Search engines start to favour AI-accessible and AI-enabled websites. Human UX still has value, but AI usability is prioritised. We may also see search engines themselves become AIA. We've already seen the early version of AI-Enabled, with Google's AI overviews and "AI Mode".

Website traffic becomes overwhelmingly dominated by AI agents.

Brand expression, perhaps surprisingly, becomes more important than ever, because it has to permeate not only visual design but also structured data passed to AIs. Brands will need to figure out how to convey their essence in the age of AI agents and intelligent things.

Who Wins

Early adopters and first-movers. Those who build for the future, rather than focusing on what's "hot" today. Platform providers (like Shopify) who make their services AI-Accessible will see significant increases in revenue through partnerships and AI-driven discovery on the new web. In general, companies that evolve their products/services into intelligent service nodes.

Who Loses

Businesses built around or dependent on display ads or traditional funnels (obviously).

"Dumb" sites. With AI agents using AIA search engines that favour AIA and AIE websites, the traditional website will be buried in results pages. The intelligent neosites will reign supreme. A million websites that fail to adapt will fade to irrelevance.

What Remains

Human-centric platforms. Forums, communities, social media, and other spaces where the value is in human connection and shared ideas and experiences. Sadly, this is likely where display advertising will also consolidate.

Gaming platforms fall under the same umbrella. But things get interesting when generative gaming becomes viable and mainstream.

Underlying infrastructure, for the most part. AI may help us improve it, but the current emergence won't fundamentally change it.

Human decision-making. The human-in-the-loop remains, but supercharged with AI power.

What Now?

It doesn't matter if I'm wrong about every single thing in this post. The one thing that is undeniable is that AI is a revolution that is changing how we do everything. So, even if things don't change the way I suspect they will, we can at least be certain of one thing: things will change.

The question is: Will you be ready? Personally, I'm already building for the future. Are you?