Beyond the Browser

I've been testing some of the new AI browsers that have launched recently, and I keep bumping into the same disappointment: after twenty years of browser stagnation, this is the best we can do?

Perplexity launched Comet. The Browser Company released Dia (and then got acquired by Atlassian). Opera announced Neon. Anthropic made a Claude plugin for Chrome. Google (finally) integrated Gemini into Chrome. Microsoft has Copilot Mode in Edge. Everyone's racing to build AI-powered browsers. But they're all doing essentially the same thing - adding a sidebar with an AI assistant that can look at your tabs and answer questions about them, or do some basic browsing and form-filling.

The problem is, these AI browsers aren't making things any simpler. They're just adding an AI layer (typically in a sidebar panel) on top of an existing mess of multiple windows and tabs (and now, split tabs). And when I need AI to synthesize information or answer a complex or challenging question, they're less capable than ChatGPT or Claude or Gemini. They're sporks - trying to be both browser and AI assistant, ending up mediocre at both.

Here we are in 2025, we finally have AI technology powerful enough to reimagine how we interact with information on the web, and all these companies have done is bolt it onto the same interface we've been using since the early 2000s. It's a bit disappointing.

The Incremental Trap

These products have added AI assistants to help navigate tabs, synthesize information across multiple windows, and find things you've lost track of in your browsing chaos. But they're not simplifying anything. They're building another layer of complexity to manage the existing complexity.

Tabs and bookmarks were attempts to manage the mess of websites to keep track of. History was supposed to help you find things. Browser sync was supposed to unify your experience across devices. Profiles were added to separate work and personal browsing. Virtually every browser innovation for the past twenty years has been another organizational tool to help humans cope with an interface that doesn't match how we think.

None of it has worked particularly well. People still have fifty tabs open. Bookmarks still pile up unused. We still can't find that article we read last month. The browser paradigm itself creates chaos faster than any organizational feature can contain it.

And now AI browsers are adding another layer. Instead of questioning whether tabs and windows and bookmarks make sense, they're building AI features to interact with your tabs and windows and bookmarks. They're not eliminating the problem, they're just offering a more sophisticated tool to cope with it.

So what do these AI browsers offer? A chatbot that can look at your tabs and try to make sense of them. An agent that can automate tasks across windows. Synthesis features that pull information from multiple sources. The ability to purchase a ticket to the Super Bowl (this took Comet a few tries but eventually it got me all the way to the checkout page. No, I did not buy one, it was just a test.)

But here's the problem: if I want a powerful AI assistant, I'll use ChatGPT or Claude or Gemini. They're better at understanding context, better at synthesizing information, and better at complex reasoning. And if I want to browse the web efficiently, I'll use Chrome or Safari, which are faster, more stable, and have decades of optimization behind them.

AI browsers currently occupy this awkward middle ground where they're not as good as dedicated AI chatbots for AI tasks, and not any better than traditional browsers for browsing tasks. They're adding basic agent capabilities to the browser instead of asking what a truly AI-native interface for accessing web information might look like.

What Could Be

Here's what's interesting: many people are already experiencing what a more ambitious approach looks like. When you have a question, you might ask ChatGPT instead of googling and clicking through search results. When you need to research something, you might have Gemini or Claude do some "deep research" to synthesize information from multiple sources instead of opening twenty tabs. When you want to find a restaurant to match specific criteria, or suggest activities for your next trip based on your needs and travel dates, you can ask an AI for personalized recommendations.

The "browser" in these interactions is invisible. There's no URL bar, no tabs, no bookmarks. You interact through conversation, and the AI handles whatever web access it needs behind the scenes. The chaos never reaches you because you never enter the browser paradigm in the first place. And it works better. You get better answers faster without the cognitive overhead of managing multiple tabs, evaluating search results, or trying to remember where you found something.

This is what makes me think OpenAI might be approaching this differently. While other companies are trying to build better browsers, OpenAI already has millions of people using ChatGPT as their go-to interface to information. They don't need to build a better browser—they're making ChatGPT better at accessing and synthesizing web information, and even taking action on web pages with Agent Mode. Taking it one step further, they just launched a new feature (U.S. only for now) called Instant Checkout that lets you purchase items from Etsy and Shopify merchants directly within the chat interface. Suddenly the question goes from "which browser?" to "do I even need a browser?"

The Opportunity We're Missing

We're at a moment where AI could fundamentally reimagine how we interact with information on the web. Not just make the current experience slightly better, but actually solve the problems that browsers have never quite been able to fix.

Imagine an interface where you describe what you're trying to accomplish, and the system figures out what information to access, what actions to take, and what to show you. No manual tab management. No bookmarking things you hope to find later. No switching between twelve different windows trying to remember which one had that important detail. Just intent and results.

This sounds radical, but it's already how we interact with other complex systems. If you want to play a specific kind of music on Spotify, you can just tell it what you want to hear and it figures out what to play. The complexity exists, but the interface hides it.

The browser could potentially work in a similar way, but it would need to be completely reconceptualized. Instead of taking that leap, these AI browsers are taking baby steps - adding AI sidebars called “assistants” to help us manage the same chaotic interface we've grown used to. It’s better than nothing, but it could be so much better.

The Web Ecosystem Question

But we need to be careful not to throw the proverbial baby out with the bathwater. The web as we know it, for all its messiness, has evolved over three decades to become an international open space for human creativity, discovery, and verified knowledge. It's a great big beautiful mess!

There are valid concerns that the web could wither if it essentially becomes content infrastructure for corporate AI gatekeepers who don't "give back" to the millions of publishers, merchants, institutions, and individuals who have built their web properties for humans, not AI agents. It's even been called an "extinction-level event" by some publishers. How do we maintain the balance of the ecosystem? I'm not sure anyone has the answer yet.

The web as it works today required a huge investment in effort and dollars that many people are understandably very interested in protecting and preserving. Completely rethinking and successfully reimplementing revenue models for a fully AI powered future will likely take some time.

Why This Evolution Matters

After twenty years of essentially the same browser experience, we finally have technology powerful enough to do something genuinely different. And the response from the industry has so far been underwhelming. They're iterating when they could be reimagining.

All of the companies building AI browsers are taking the safe bet. They're trying to make the browser work better by adding AI, when there's an opportunity to ask much bigger questions about what comes next.

This isn't to say these AI browsers serve no purpose. They represent evolutionary progress, and evolution matters. But evolution is what you do when you're refining something that already works well. When something has been fundamentally flawed for twenty years, you need revolution, not refinement.

Maybe that's why OpenAI is rumoured to be working on a browser but hasn't announced anything. Maybe they realized that what they need to build isn't a browser at all. Maybe they figured out that they already have a post-browser interface, and millions of people are already using it every day without thinking of it as a browser replacement. We'll see.

Ultimately it won't matter much which AI browser becomes the most popular. What we need to watch for is how often we find ourselves bypassing the browser entirely, working directly with our favourite AI chatbot to accomplish what we used to need tabs and bookmarks and navigation for. Once that becomes the default instead of the exception, we'll know someone finally made the leap these AI browsers are missing.

Cover image generated with Midjourney. Research and editing assistance provided by Claude Sonnet 4.5