What Happens to Websites When AI Does the Browsing?

When you last searched for a restaurant, did you actually visit the restaurant's website? Or did you ask something, get a summary with ratings and hours and a menu excerpt, and only click through when you were nearly certain?

That behavior has been shifting gradually for years. AI is about to accelerate it into something far more structural.

Claude, Gemini, Perplexity, and others can now search the web, read content, compare options, and in some cases take action on your behalf. Book the table. Find the flight. Pull the pricing. Summarize the reviews. The user never has to visit the source. The AI does the visiting.

So here is the question I've been sitting with: if the AI is doing the browsing, what is the website actually for?

The Database Part of the Web Is Being Quietly Abstracted Away

Most of what we call "websites" are actually structured data wrapped in a presentation layer. A restaurant's website is mostly a database: name, address, hours, menu, price range, reservation link. A hotel booking platform is a database with a calendar attached. A recipe site is a database of ingredients, timing, and steps.

For all of these, the "page" was always an intermediary. We built pages because browsers needed somewhere to go, and users needed somewhere to look. But if an AI agent can extract that underlying data directly, the page becomes optional. Not wrong. Not dead. Optional.

This is already happening. These AI tools are pulling structured information from across the web and presenting it without requiring you to visit the source. For the database-heavy part of the web, the website-as-destination model is under real pressure.

What This Means If You're Building Something

Here is where it gets uncomfortable for product builders. If all your product contains is data that an AI can read and relay, you don't have a product. You have a data source.

That's not a criticism. Being a good data source, well-structured, accurate, consistently maintained, is genuinely valuable. The web will need reliable data more than ever. But the presentation layer sitting on top of that data is no longer the competitive advantage it once was.

The builders who understand this will stop spending most of their energy on how the page looks and start thinking about the underlying structure. Is the data readable? Queryable? Well-organized? Can an AI extract what matters from it, and extract it correctly? These questions are becoming at least as important as whether the homepage converts.

This is also why I keep returning to the distinction between database products and experiential products. The database part of the web, pricing, availability, schedules, menus, directories, is being abstracted away. The intermediary layer is disappearing. But the experiential part is not.

What Survives the AI Browsing Era

Here is what I believe cannot be abstracted away: the experiential layer. The part of the web that earns attention not because it contains information, but because of how it presents it. Voice, perspective, editorial judgment, design that creates feeling rather than just organizing data.

A personal newsletter is not a database. A brand story is not a database. A piece of writing that shifts how you think about something is not a database. An AI agent cannot surface these things the same way it surfaces opening hours, because they are not primarily containers of information. They are experiences. They require a human layer to exist.

The question for anyone building digital products right now is not "will websites survive." It is: which part of my product is database, and which part is experience?

The database part will increasingly be accessed without the page. Build it like a database: structured, queryable, accurate, ideally accessible via API.

The experiential part still needs the human layer. The voice, the judgment, the editorial decisions: these are what make someone choose you over a hundred AI-surfaced alternatives. Build those like a craft, not like infrastructure.

The Page as Printed Brochure

My guess is that in ten years, the general-purpose marketing website will be what the printed brochure is today: useful for some things, nostalgic for others, no longer the default.

The web is not going away. But the page as the unit of the web is being renegotiated. For the database part, structured data that AI can read replaces it. For the experiential part, pages will likely still exist, but they will need to know they are competing in an environment where information is already abundant.

The builders who do well in this environment are not the ones who fight the shift. They are the ones who are honest about what kind of product they are building, and design it accordingly.

I'm still working out what this means in practice. But the question itself, what is this product for in a world where AI is doing the browsing, has become the most important thing I ask at the start of any project.