The internet is about to flip — and most of us aren’t ready
I went down a rabbit hole today looking at technology adoption data and honestly it kind of blew my mind.
The telephone took 75 years to reach 100 million users. Electricity — 46. Radio — 38. TV — 26. The internet — 7 years. ChatGPT did it in 2 months.
Two months.
But then here’s the weird part — only 16% of the world has ever touched a generative AI tool. Even once. In Africa 62% of people don’t have internet at all. In Japan — Japan! one of the most technologically advanced countries — AI adoption is just 17%.
So we’re living in this strange moment where it’s simultaneously the fastest technology adoption in human history and something most humans haven’t tried yet.
I was watching a talk by Fabricio Bloisi, CEO of Prosus (they run iFood, OLX, Takealot and a bunch of other big platforms). At an AI conference in Lisbon he said something that stuck with me — that very soon websites will be visited mostly by bots, not by actual users. Especially marketplaces. His company already has 30,000 AI agents running.
And when you look at the numbers he’s not wrong. Bot traffic passed human traffic for the first time in 2024 — 51% of all web traffic is now bots. By the end of 2025 there was 1 AI bot visit for every 31 human visits. At the start of that same year it was 1 per 200. That’s insane growth. And human traffic actually dropped 5% in one quarter.
So his conclusion is as a business you should make your website more accessible for robots and agents, not just focus on UI/UX for humans.
But I’m not 100% agreeing with him. Or rather I agree with the direction but not the timing.
We’re not there yet. Not for normal people.
There’s a massive gap between “I asked ChatGPT to help me write an email” and “my AI agent browses marketplaces, compares prices, and buys stuff for me while I sleep.” That second thing requires trust. Real trust. Like letting-someone-else-use-your-credit-card kind of trust.
And if you’re living on a tight budget — in Lagos, São Paulo, or even here in Lisbon — one wrong purchase by an AI agent is not a fun experiment. It’s a problem.
Technology is ready. Trust isn’t. That’s the actual bottleneck.
But then again… “not yet” in tech years could mean 18 months. The whole history of technology adoption shows that each wave rides on the infrastructure of the previous one. AI doesn’t need new hardware — 5 billion people already have the device in their pocket. It just needs a software layer on top.
Remember how Telegram went from “what is this” to “my grandmother sends voice messages” basically overnight? The same thing could happen with AI agents. Someone packages it into something dead simple, and then it’s a wave.
OK so what does this actually mean if you’re building something?
I think you need to run two tracks at the same time. You still need great UI for the billions of people who browse, click, scroll and buy manually. That’s not going away tomorrow. But you also need clean APIs, structured data, machine-readable product information — because the bots are already here and they’re growing fast.
Go all-in on agents too early and you lose 90% of your users who still need a button to click. Ignore agents completely and you’ll lose the growing automated traffic.
And now the part for my fellow frontend engineers.
We spent years — YEARS — perfecting pixel-perfect hover states. Debating whether border-radius should be 8px or 12px. Fighting about padding values. Losing sleep over that one animation that stutters on Safari.
And now the primary visitor to our websites might not have eyes.
Your beautiful gradient? A bot doesn’t care. That buttery 60fps scroll animation you spent two days on? Completely wasted on a crawler. The great light mode vs dark mode debate? An API call has no preference.
I’m joking. Mostly.
The reality is this doesn’t kill what we do — it splits it. There’s still the human-facing layer and it still matters a lot. People need to trust what they see, understand what they’re doing, enjoy the experience. That’s not going anywhere.
But now there’s a second layer that matters just as much — the machine-facing one. Structured data. Semantic markup. Clean APIs. Agent-friendly auth.
The frontend engineers who do well in the next few years won’t just build interfaces for people. They’ll build for people AND their AI assistants at the same time.
Or to put it differently — you’re not losing users. You’re gaining a second audience. One that doesn’t care about your font choices at all but has very strong opinions about your data structure.
The internet was built for humans to read. It’s becoming a place where machines read it on behalf of humans.
That’s not a bad thing. It might even be the thing that closes the digital divide — if someone in rural Nigeria can tell their phone “find me the cheapest laptop with these specs” and an agent does the rest, they don’t need to know how to navigate Amazon’s UI. They don’t even need a laptop.
Or it could widen the divide further. That part is up to us.
75 years → 46 → 38 → 26 → 7 → 2 months.
The curve only goes one direction.
This text was written with the help of my favourite AI model.