When the web browses itself

Human Web Browser is the most endangered job due to AI

Bots Eat The Web, concept by Baratunde and Layne Deyling Cherland. Visuals by ChatGPT & Unacknowledged Human Artists.

Hi you,

For thirty years the web has been a glass box made for human eyes—pixels, banners, heat-maps, clicks—all premised on you sitting in the driver’s seat. That era is coming to an end. The next web is designed for software. Think of asking, “Plan me a three-day craft cocktail crawl in Portland, Oregon,” and your personal agent returns a custom microsite, turn-by-turn audio walking tour directions as podcast, and a share-ready map before you’ve finished your morning espresso. In this “bespoke web,” bots—not people—do the browsing, assemble the answers, and negotiate the tolls. It’s the end of the web as we know it, and it’s not all good.

The Job Really Affected by AI

It isn’t the truck driver or the lawyer whose labor will be most affected by deploying AI; it’s the human web browser. The job of opening tabs, skimming pages, and pasting links, may soon be behind us meat-beings. That’s not necessarily a good thing. There’s definitely some value in, I don’t know, actually processing information yourself. After all, how are extremely online people gonna tell people to Do Your Own Research if all the research is done by machines??

Last year, The Verge’s Nilay Patel coined the term “Google Zero” for the moment Google stops sending traffic to websites in favor of its own AI-generated answers to search queries. Now, some outlets are already seeing double-digit referral drops. That’s why Cloudflare’s new Pay-Per-Crawl program matters. It allows publishers to charge AI agents per request instead of hoping for human ad clicks to come back.

It’s not just that website traffic from humans is dropping. It’s that traffic from robots is exploding. Google and search engines have had web crawlers for a while. For every two robot scrapes, Google would send one human visitor. With the latest AI frenzy, those ratios are exploding. For Google it’s now 14-to-1, robots to humans; OpenAI is 1,700-to-1; Anthropic, founded by people who left OpenAI because they felt that company was too reckless, comes in at 73,000-to-1. Not a typo. For every seventy three thousand times Anthropic sends a robot to a website, they send one human. Ratios like this are what we might call unfair trade. Maybe this is where tariffs are a good thing?

Death (or Mutation) of the Page

I don’t think the webpage is some inherently, forever-valuable thing worth defending at all costs. Some of my worst experiences were brought to me by a webpage (I’m still traumatized by pop-ups and pop-unders). I don’t feel beholden to the stone tablet, though that did force you to really really think about what you wanted to write because corrections could take a lifetime. My main point is there are other ways to disseminate information. In a post-human-browser world, something else is coming. Traditional pages fade as agents compile just-in-time, just-for-you bundles: NotebookLM spits out an audio brief; Perplexity builds a mini-wiki; Meta AR glasses layer a 3-D overlay that only shows you videos of Mark Zuckerberg on his airfoil singing God Bless America while rocking a MAGA hat. It’s called INNOVATION. The result is that websites are going away, replaced by something that looks more like an API endpoint than a destination.

For an idea of how this might show up on an e-commerce property formerly known as a website, check out my conversation with Bloomreach CEO Raj De Datta.

Follow the Money and the Metrics

The business model flips from advertising to humans to metering the machines. Time-on-site and CPMs give way to requests-per-minute and tokens burned. But in this brave new world, what keeps creators compensated and credited? Who’s going to have the economic incentive to produce information in the first place?

I recently met the founder of Credtent which licenses artist data to LLMs, and is prototyping a form of bot-to-creator payment that could restore some balance to the force. Stay tuned here for a future conversation with them about how it works.

The Age of Ultra-Processed Information

But what about the information ecosystem? If you still believe in things like facts and objective reality or at least generally-shared reality, then you’re in for some… adjustments. When bots copy from bots, errors don’t just spread—they compound. We are creating a world in which we are increasingly abstracted away from information sources, kind of like our food system. We risk entering an age of “ultra-processed information” made of harmful additives and empty calories that does harm to our societal bodies.

For an insightful and emotional exploration of some of what a world stripped of objective reality looks like, check out my conversation with Witness’s Sam Gregory.

In response we’ll have to see the rise of tech and standards that can prove the provenance of information—an information supply chain verification scheme, a robot.txt for payments, authenticity headers, watermarking—as quickly as agent traffic grows.

---

Why all this Matters to our Life With Machines

• Credit & Compensation – The incentive to publish collapses unless we invent bot-era royalty systems.

• Information Integrity – Without transparent sourcing, we’ll drown in hallucinated summaries that outrank the originals.

• Civic Power – If a handful of agent-platforms mediate reality, that could be a wrap on self-governance since an informed public is a pre-requisite for democracy. Are you having fun yet??

---

One Big Question for You

How would you price your knowledge for a world where machines, not humans, are the primary visitors?

Hit reply. Your smartest takes (or funniest rants) could land in Thursday’s digest.

— Baratunde

Thanks to Associate Producer Layne Deyling Cherland for editorial and production support and to my executive assistant Mae Abellanosa.

Here’s an unedited voice memo of the thoughts that drove this newsletter, just to let you in to the process a bit more, and because I like actually speaking.