A picture of a computer lab. Source from Wikimedia Commons.
If you search "bring back the old internet" today, you'll find thousands of people saying it. From essays to comment sections and even on the platforms they're complaining about. The phrase entered wide circulation around 2020 with multiple mainstream online writers focusing on the idea that the internet changed for the worse. What people mean by the sentiment revolves around the viewpoint that the web from the late 1990s through the 2000s was a freer and more interesting place. That era was filled with independent message boards, personal websites, blogs you found through other blogs, and search engines whose whole function was sending you somewhere else. This is different from what the first generation of internet users mean when they talk about the "old internet." For them, the real "old internet" ended in September 1993, when AOL opened Usenet to its subscribers and the annual wave of newcomers never stopped. Usenet regulars named it Eternal September and have been mourning ever since. And indeed, there is a huge difference between the internet before 1993 and what came afterwards. But that difference deserves its own focus. This article, however, is about the later era that most people today mean by the "old internet." The current argument predates the slogan that started popping up in 2020. In December 2012, Anil Dash, tech entrepreneur and former advisor to the White House Office of Digital Startegy, published "The Web We Lost." In it, he pointed out that users had once been able to export their data in full, move it between competing services, and own their identity at their own domain rather than as a handle appended to someone else's company. It drew a huge response and led to a presentation he gave at Harvard's Berkman Center. Cory Doctorow, blogger and science-fiction author, later gave the new internet a name that stuck: enshittification. It was the American Dialect Society's 2023 Word of the Year, and it revolved around the tendency for two-sided platforms to decrease in quality over time because the focus moves from the content of the platform to the business interests that the platform serves.
What people are objecting to now about the internet includes things that the dominant platforms like Facebook and Google view as positive: algorithmic feeds that push content, engagement optimization, and the consolidation of content in one site. The most extreme formulation is the so-called dead internet theory, which seems to have originated in an online post in January 2021 and became popularized by The Atlantic's coverage of it in that magazine's September 2021 issue written by Kaitlyn Tiffany. In essence, the theory is built around the idea that the web is now mostly bots and AI content coordinated by governments and media to manipulate the public. And that the change to the new, artificially generated internet happened sometime around 2016 or 2017. It obviously has elements that ring true in terms of the rise of artificial content. A good example of mass generated content is what happened during the 2016 Presidential Election in the United States and the counterfeit and false news posts generated by political and national actors, to include by Russia's Internet Research Agency. But there's even more to support the idea that something changed in the late 2010s, even if it doesn't necessarily validate all of the dead internet theory.
Arbor Networks found the top 150 content and service providers carried 26% of all internet traffic in July 2007. But by 2009, they held 52%. Half of all internet traffic went from being generated by five to ten thousand companies to being generated by mostly just 150 companies in twenty-four months.
The bedrock of internet publishing itself became a moving target in the 2010s. Hypertext Markup Language (HTML) and Cascading Stylesheets (CSS) ceased to be specifications with distinct versions and instead became "living standards." For its first two decades, the web had versions. HTML was standardized by the World Wide Web Consortium, which was the body Tim Berners-Lee founded in 1994. It published HTML 3.2 in 1997, HTML 4.01 in 1999, and CSS 2.1 in 2011. Each version was a dated document with a fixed feature set and was published after a formal review process. You could read one, implement it, and know when you would have ample heads up if anything changed. But around 2012, HTML became constantly evolving with the roll out of HTML5. The practical consequence was the end of independent browser engines since the time and resources needed to stay current with the changes rolling out by the web giants like Google every few months priced out only but the biggest players. Opera abandoned Presto for Chromium and Mozilla's Servo effort was gutted in 2020. Building a browser stopped being difficult and became nearly impossible for anyone without a thousand engineers. This now means one company now effectively defines what the web is. In the same window, Twitter imposed an algorithmic timeline in February 2016 and Instagram abandoned chronological order in March of that year. And Google launched AMP, which involves basically serving other people's pages from its own servers. Deep-learning recommendation got much better and mobile usage passed desktop globally. Selection replaced navigation. And it became harder to chose what you saw after that.
Perhaps a better way of framing the situation is through absence. The internet that we knew for decades is literally disappearing. Thousands of independent forums, each small enough that reputation was tracked by people rather than metrics, have shut down. Online advertising, which allowed independent creators to make a living has been centralized into only a few remaining options with Google's Adsense the dominant player. And instead of a delivery system that delivers content without question, the Internet has become discretionary. Social platforms like YouTube demonstrate the problem clearly through their almost daily takedowns of channels and demonitization. And the discovery aspect--which was the driving force behind the internet's structure--has since vanished. Websites like StumbleUpon and del.icio.us, which were built to provide new content with little restriction and no algorithmic pushes are gone. Even basic directory sites are closing, which makes finding an address for something more and more contingent on proprietary search or AI platforms. So what caused all this?
Law shaped a great deal of this new internet. CAN-SPAM, passed in 2003 to stop junk email, preempted stricter state laws but also shifted the real power to the filters. This is something I view as the start of the "discretionary internet." The stated reason for this was to protect from spam. But having a third party decide what gets through was not the only way to solve this. There were those who said the best solution was to allow users to decide themselves what sender could message them again. Over two decades, this has resulted in a handful of providers now deciding which messages arrive. And the discretionary model has spread to now include text messages over phones and messages via social network platforms. DMCA Section 1201 and the Computer Fraud and Abuse Act made interoperability legally hazardous, which is why no one can build a third-party client for anything. FOSTA-SESTA in 2018 took out Craigslist personals and Tumblr. And Britain's Online Safety Act now covers an estimated 100,000 services under penalties reaching up to millions of pounds. The result is that the cost of compliance and pentalties for violations have pushed independent voices out while allowing the big players with the money to pay the costs to not only stay but get even bigger.
HTTP status code 402 for "Payment Required" has been in the web's specification since 1997 but was never implemented. With no native way to move small sums for small content, advertising became the only workable model.
The economics finished it. Venture capital stopped funding many small companies and started funding a few enormous ones. US venture hit over $400 billion in the first half of 2026 with AI capturing 86 cents of every dollar. And two companies, OpenAI and Anthropic, are estimated to have taken around 40% of all startup funding worldwide. And the independent, innovative companies that do get venture capital or funding from bigger companies are often shuttered with the valuable pieces going into someone else's platform. Underneath that, everything became a rental. Software became subscription. Servers became cloud accounts. Music and film fragmented into a dozen streaming platforms where nothing is owned and titles vanish without notice. Your audience became a database entry on someone else's platform. It's non-portable by design. The old internet's real advantage was never its aesthetics. It was that leaving was cheap, message delivery wasn't discretionary, and you had many more options to make money without having to rely on a giant corporate platform like YouTube. That's the new internet and why people want to bring back the old one.
