The Internet is Closing
Author: Sameer Vijay, with input from Daniel Porras Reyes and Chip Hazard
For decades, we built on an open web, where anyone could publish, Google indexed everything, browsers offered free access, and monetization flowed through search and social. This fueled massive industries like programmatic ads and digital marketplaces.
That ecosystem is now fracturing. AI chatbots like ChatGPT have created a new paradigm. People no longer need to browse blue links or scroll feeds to find what they need. They ask and receive answers.
This shift has enormous implications. The open internet that enabled ChatGPT, with its vast corpus of freely accessible training data, is now being disrupted by the AI it created. As the internet closes, traditional methods of monetizing attention through SEO, search ads, and display ads are breaking down. We are at the start of a generational platform shift. As the old rails of digital marketing erode, a new system is emerging where data is proprietary and AI directs attention.
How Protocols Built the Early Internet
The web began as an open set of protocols. While telecoms and tech giants like Microsoft debated who would build the “information superhighway,” pioneers like Marc Andreessen saw it as a bottoms-up effort when he built Mosaic. With open protocols, based on Tim Berners-Lee’s work, like URL, HTTP on TCP/IP, and HTML, anyone could create a website if they could host it. As individual sites came online, they formed a dynamic, complex web. It was a patchwork network built bottoms-up because the internet was fundamentally open.
That openness had major implications. Some products became essential for individuals to get value from the internet. Browsers enabled people to load and navigate sites. Search engines indexed the network and helped users find sites by keywords. Advertising became the default business model. A free, open internet attracted more users, and ads best monetized that attention. The link economy emerged as people visited sites, generating ad revenue for bloggers, forums, and niche media. Publishers created streams of data that made the open internet engaging.
An oversimplified version of the original advertising value chain on the open internet looked like this:
Publishers created websites and content to attract users, embedding ad spaces (ad supply) to monetize traffic.
Users went to Google to connect with websites.
Google decided which websites to send users to, with search ads helping brands gain traffic.
Advertisers/brands paid to reach users through publishers’ ad spaces (ad demand) and Google search ads.
Ad networks connected supply from publishers and demand from advertisers, taking a cut as brokers.
What Happens When the Web Stops Linking?
The open internet worked because search engines drove traffic to publishers. Now, answer engines are driving very little traffic to publishers:
The percentage of users who clicked on a Google search result used to be 15%. With an AI Overview present, that number dropped to 8% (Pew).
The percentage of a publisher’s traffic coming from a Google referral was 91% in Q2 2024. In Q1 2025, that had dropped to 85%, likely due to AI Overviews (Tollbit).
Bing’s scrape-to-referral ratio is 11:1 – for every 11 scrapes, Bing returns one human visit to a website. OpenAI’s scrape-to-referral ratio is 2k:1, Perplexity’s is 202:1, and Anthropic’s is 71k:1. (Cloudflare, Tollbit)
The link click-through rate for answer engines is 0.67%, compared to 8% overall for Google search.
Click-through rate for AI apps (answer engines) vs top 10 Google search results (Tollbit)
The likely reason for this drop in traffic is that answer engines remove the need for users to click links to get information. The “original sin” of LLMs training on internet data for free came from the internet’s fundamental openness.
What Comes Next?
As the internet shifts from open to closed, we believe we are seeing a generational platform reset. It mirrors the rise of the open web in the 2000s, when entire industries emerged to help businesses monetize, discover, distribute, and measure. This time, the focus won’t be on monetizing attention through ads but on monetizing data through licensing and subscriptions. That shift breaks the traditional internet value chain and creates space for startups to build the infrastructure the next wave of the web will depend on.
We see five key opportunities for founders:
Build marketplaces to connect the long tail of data creators with AI labs hungry for fresh, proprietary content.
Develop data valuation frameworks to price content in a world where impressions and clicks no longer suffice.
Create bundling and personalization infrastructure to turn fragmented niche content into high-value subscriptions.
Build identity and permissions tools to help publishers control access and distinguish between bots and humans.
Design data rails, structured LLM-ready publishing formats that let content flow cleanly and securely from creators to AI systems.
We believe this new wave of startups will form the foundational systems of a new internet era. The monetization engines, discovery layers, and publishing protocols of a closed web. Founders who move early will shape the next generation of infrastructure and own a vital piece of the LLM economy. If you are building in these spaces, we want to hear from you.
