the internet is dead. long live the internet.
The infamous Dead Internet Theory has been growing more and more real (and costly) by the day. And experts say that in two years, it will be fact.
Perhaps one of the strangest theories about the web is that what we think of as a very large, active internet being used by billions of people worldwide, is actually anything but. Instead, almost all activity online is bots, scripts, and hypervisor software talking to each other, and human interactions are just a small surface layer. If you’re a little bit more conspiracy minded, add-ons to the theory say that all this fake traffic is used to manipulate search results to trick consumers and spread government propaganda.
Why is this idea so strange? Because it’s being treated as a weird conspiracy theory while it seems to be careening towards being true. In fact, as a professional techie, I would even go as far as wholeheartedly endorse it despite some major media outlets dismissing it as a paranoid fantasy while still confirming large parts of its premise.
Take email for example. Every day, a third of a trillion — yes trillion, as in a thousand billion, you read that right — emails are sent. Just shy of half of those are spam, down from their all time peak of 56.6% in 2017, although total numbers have increased. Of the non-spam emails, too many are marketing messages from companies who think that one purchase now means you’re forever pen pals, notifications from financial institutions, service providers, and newsletters. (Hey, and this is one of those very newsletters! How meta, huh?)
Okay, let’s move on to websites. There are just a smidgen over a billion of those if we look across the web, roughly half of which are WordPress instances, which is why the programming world can’t rid itself of the scourge that is PHP. Most will only get a few hundred clicks a month but almost half of that, or 47.4% of all web traffic, comes from a bot or script. Almost a third of those are ad and click fraud bots, and the remainder are crawlers for search engines, AI training scrapers, business automations, and data gathering for metrics and research.
wait, what do humans actually do online?
Of the platforms where we know flesh and blood humans do spend enough time, 92% of all content is created by 10% of all users, and engagement with different posts can vary from 0.03% to 0.1% of all viewers. In other words, of the 52.6% of internet traffic which is human-driven, about 9 in 10 users stuck to messaging friends and family and just passively consume content mostly meant to sell them stuff, or see if they would be interested in propaganda campaigns ran by governments, both theirs and foreign, often using bots and fake accounts to boost their messaging.
And now, AI is threatening to make things even worse. We’ve previously talked about the potential for chat bots to flood the web in nonsense that will immerse people in their chosen hyper-partisan reality from a rapid-fire spam-a-ganda cannon, or flood their screens and inboxes with ads. According to the latest estimates from Europol, every one of those concerns seems to be coming true and by the end of 2026, AI-generated content will account for 90% of everything we see online given its current trajectory. Meanwhile, bot traffic is growing by around 5% per year.
At this rate, not only will 9 in 10 pieces of content on the web be fake in two years, but 99.96% of all traffic to this content will come from bots and crawlers by 2037 if these bots aren’t forcefully reigned in. This doesn’t mean no one will use social media, read the news, or message their friends. It’s just that they will be overwhelmed by millions of bots, scrapers, fraudulent click scripts, and background services meant to handle and mitigate all that non-human traffic.
The bottom line here is that maybe, the internet isn’t completely dead yet depending on your definition, but it’s definitely careening towards death at an accelerating rate. Human traffic is plummeting and our engagement shows passive consumption being the default. AI-generated traffic, bots, and scripts are growing at alarming rates and are threatening to overtake human content in the matter of a few years. By that time, bot traffic will surpass that of humans 57% to 43%, and by the end of the decade, as much two thirds of all web activity will be non-human in nature.
the staggering costs of a dying internet
So, at which point will the internet be officially dead? Nobody really knows. What we can agree on is that automated traffic and auto-generated content is both rapidly and exponentially overtaking human activity, and that much of the bot activity is used for defrauding ad networks, crawling and reindexing the aforementioned billion websites for search engines, chatbot training, and generating fake traffic and engagement for white collar crime and propaganda campaigns.
Oh, that’s right, I forgot to mention that a great deal of that automated traffic costs a lot of companies tens of billions of dollars per year in fake clicks. Last year, the tally was almost $36 billion, on par with the GDP of Macau. And yes, of course it’s getting worse with AI. Even before the pandemic one in five ad clicks and impressions have almost certainly been fake. Today, it’s closer to one in three, with half of all ads never even seen by a human.
As for internet scams, Americans alone lose over $10 billion a year to online criminals. Global figures are hard to come by because fraud is often severely underreported for a variety of reasons — mostly thanks to some form of embarrassment or denial — but figures range around $55 billion per year, more than the entire economy of Uganda.
From dating and romance scams, to crypto Ponzi schemes, to phishing bots on dating sites, to a nasty ransomware infection, to good, old fashioned take-the-money-and-run fake storefronts, there are now more ways than ever for internet bots to raid your wallet. That means companies and individuals are losing the GDP of a middle income country every year to software designed to convince you to give someone money to steal and embezzle it, or believe someone’s astroturfed propaganda.
All of that sure makes the Dead Internet Theory sound more like a quick glimpse into the very near future rather than a paranoid fantasy, and all of us who work with cloud tools, AI, and the web agree that we haven’t even seen close to the worst of it.
In a way, it was always expected that machines would take over the internet. Most of us look at the surface web, which by now is five companies in a trench coat — Meta, The Platform Formerly Known As Twitter, Google, Amazon, and Microsoft — and see its collection of sites and platforms as, well, the entire digital universe. But really, the internet is a collection of very useful standard protocols like UDP, TCP/IP, SSH, HTTP, and — a personal favorite for some of my projects — gPRC. It’s a no brainer to make machines talk to each other with one or more of the ingredients in that alphabet soup.
Likewise, the more devices there are out there, the more connected they are, and the farther away they are, the more traffic across the infrastructure of the web will carry pings, acknowledgments, status updates, and messages in a work queue. Any human browsing a website or refreshing a screen in an app is not going to generate nearly as much activity by definition, as software operates on a nanosecond scale while we are literally a few billion times slower.
what we lose when the human internet dies
And yet, we comp sci professionals hoped that this tsunami of machine traffic using the underlying internet infrastructure will be, well, useful. Like messages to make life more convenient through the Internet of Things, or help with realtime navigation and tracking important things in meatspace. To be fair, all that exists and is out there now, being used by humans to get tangible, real world benefits. But even all that useful automated traffic is rapidly losing ground to click, crime, and scam bots that provide the exact opposite of tangible, real wold benefits.
Most of this is thanks to what’s known as the Original Sin of the Web. To get people to use the internet, most website creators decided to make access free and fund all their work by advertising. At the time, this seemed like a good idea because that’s how you moved zines and no one was going to pay for access to one of 30,000 websites which mostly looked like this. Not knowing much about the web, but knowing that the work was done by humans on a mission, advertisers paid well, and a lot of sites got off to a smooth, profitable start.
Unfortunately, as we became more and more immune to advertising and companies couldn’t interrupt our experience by showing those ads in our face, they got metrics to strong-arm websites into ever lower payouts. This encouraged ever more click fraud, especially after advertisers offloaded the mechanics of ad delivery to networks like Google, Ad Wave, and Tribal Fusion, which are trivially tricked into funding scams, piracy, snake oil, and disinformation, pushing advertisers to get more desperate and more intrusive until sites now look like this without an ad blocker. Oh, and all those nonstop ads also spread viruses now. Awesomesauce.
Unless we figure out how to do an ad-ectomy from the web and focus on search and AI being actually useful tools for humans rather than simply generate Content Brand Content™ for ad impressions and affiliate fees, the internet will indeed die for anyone who isn’t sticking to news aggregators, social media, and the associated apps. From the original vision of a system to share the world’s knowledge and collaborate across oceans in real time, to a digital sewer of scams, disinformation, and random bytes not even meant for human eyes, it’s a long, sad way to fall, isn’t it?
After all, what use is a hub for sharing knowledge, fun, ideas, experiences, and tools to make life easier and more organized when so much of it is floating on a vast ocean of scams designed to trick, scare, manipulate, disinform, or sell to victims of deeply rooted and reinforced confirmation bias? Why use something that rewards a nonstop steam of outrage, fraud, lies, and money laundering between crime bots, where you are little more than a necessary evil to justify a transaction or a mark for criminals and propagandists?
When the internet does officially die sometime between 2033 and 2036 if we maintain our current course, we would have done the equivalent of turning what was meant to be a college town brimming with libraries and fun or useful stores, clubs, and arenas, and turned it into a dump ran by roving gangs of con men, vicious cults, pickpockets, and dealers on corners selling Cthulhu knows what. Meanwhile, we’d be relegated to gated communities where we’re constantly bilked for every spare penny. And at that point, we may as well just start again with a clean slate…