The dead Internet: what AI regulators and you must know
Alfredo González Briseño
9/18/202611 min read
The "death" of the Internet. I read about it a few years ago. In a nutshell, the article was predicting that at some point most of online content will be created by artificial intelligence. Today this idea seems to be framed as the DIT or "Dead Internet Theory." However, this theory is not far from real as human-created content is significantly losing its share in the web. You may think this is not very important. But there are many angles to this story. Angles that governments and regulators cannot overlook. Angles that humans need to be aware of, as we spend a lot of our time online, every single day. Let's talk about it.
What is happening online because of artificial intelligence?
Far from "dead"
Of course the Internet is far from "dead." It is just a metaphor. Currently, more than 6 billion people in the world go online, according to the International Telecommunications Unit (ITU). This accounts for almost three quarters of the global population. The number of internet users increased quickly—ten years ago, there were 3 billion users or 40% of the people in the world.
According to Manochi's DataReportal, 96 out of 100 people use a mobile phone to go online. While there, Statista reports that people spend on the web between 6 to 7 hours per day, on average. A Fortune article did the math and showed that someone using the internet at this rate between ages 18 and 80, would have spent 17 years of their adult life online—the equivalent of their entire childhood and teen years.
If people almost live online nowadays, why are there conversations about the "Dead Internet"?
An article by the University of New South Wales in Australia explains that online content—including social media—is currently mostly generated by bots and AI agents. On top of that, the article's authors explain that interaction on that "synthetic" content is also predominantly done by AI. They say: "This creates a vicious cycle of artificial engagement, one that has no clear agenda and no longer involves humans at all."
How big is the AI take over?
The 2026 State of AI Traffic & Cyberthreat Benchmark Report by Human Security shows that for the first time automated traffic from software systems, including bots and AI, has surpassed human internet traffic. NBC News shared this past June that Cloudflare reported that 57.4% of requests to a selected number of websites it hosts came from automated bots, while 42.6% came from humans.
The takeover by bots, crawlers, AI agents and agentic browsers will continue to grow. In its report, Human Security mentions that automated traffic is growing eight times faster than human traffic, and that traffic from AI agents and agentic browsers grew 7,851%, year over year.
A very recent study by the Pew Research Center, "How Much of the Internet Is Written With AI?", reports that 10% of webpages show a significant sign of AI authorship. When looking only at webpages created after ChatGPT's public launch, the share grows to 35%—showing a rising trend of AI authorship online.
Behind these numbers, there is a high level of convenience many of us are experiencing when going online. An example is the AI summaries we now get when we google something. Instead of visiting multiple sites trying to find what we are looking for, we get a nice snippet generated by AI. Similarly, 4 out of 10 LLM users start their online searches using AI instead of traditional search engines, as reported by a Search Engine Land article. Their reasons: faster, clearer and less cluttered answers.
On the surface, there is nothing wrong with convenience and the shifting trends on the internet. Especially if you are a user. Yet, digging a little deeper there are some implications governments, their regulators and us in the general public need to be aware of.
Implications from the convenience AI gives us everyday
Disruptive technologies, as their adjective directly alludes, change the state of things, quickly and in a significant way. Many dynamics in the Internet—including several business models created around it—, are grounded on the basic idea that humans will consume and interact with online content.
Just think about the example of influencers in social media. Their online activities are under the assumption that people will follow them and then watch and react on their content—including influence on behavior, opinions and purchasing of products and services.
What happens to many online content creators—including social media influencers—when there is a new artificial kid on the block, who now drives engagement, becomes the main consumer of content, and even uses the human-created content, to generate its "own" content and responses?
Less organic traffic to websites
Significantly less organic human visits or clicks to websites are the immediate consequence of AI giving you overviews and summaries—when using search engines like Google—, or faster, clearer and less cluttered answers when you use Gemini, Claude or ChatGPT.
How much less? Researchers from the Indian School of Business and Carnegie Mellon University conducted a randomized field experiment and found that Google's AI Overviews reduced organic clicks by almost 40%, while increasing zero-click searches (when users stay only with the AI Overview) by almost 35%. Links shared in the AI Overview are barely clicked. Shifting from the AI Overview to Gemini-powered conversational interface in Google reduced even more traffic to downstream websites.
Why is this situation a potential issue? Leaving aside internet's purpose of sharing information, at some point the web evolved around business models based on advertising, capturing users' attention and keeping track of their online activity and engagement.
You have heard phrase like "if something online is free—including apps and information—, then you're the product." However. all of a sudden bots, AI, AI agents and crawlers get in between and disrupt this business model of content creators. An article in FastCompany couldn't put it in better words:
"Agents have no eyeballs. They do not linger over a banner ad, feel a flicker of desire at a retargeted sneaker, or click on sponsored content out of idle curiosity. You cannot surveil an agent into an impulse purchase. Every dollar of the attention economy assumed a distractible primate on the other end of the connection. The primates are leaving the building and sending their software instead."
I have not seen yet an analysis showing the economic impact of all this. However, The Gardian reported last year that sites that used to rank first in online searches dropped their incoming traffic by 79% because of Google Overviews.
IP and copyright issues
Online searches done with AI as well as AI Overviews from search engines also raise issues of intellectual property and copyright. Let's take for example your searches using ChatGPT.
When you ask the LLM a question, its "magical" faster, cleaner and less cluttered answers are only possible—I will highlight ONLY—because of the data available to the model. Part of this data comes from the data used to train it. Other parts come from the model searching answers online by using its enormous computing capacity.
The problem comes when that data belongs to creators who have a copyright on it. A peculiar ambiguity arises when people use the word "data," which sounds very technical and abstract. However, data can be a book, your article, the content of your social media posts, music in digital format, pictures, text and many other things.
To give a concrete example. A few years ago, I bought and read the book "Building a Story Brand." I really liked it and wanted to apply its ideas to my startup's website. I created my own version of the Story Brand following the book, but decided to ask ChatGPT for help. Of course, GPT generated ideas and content based on the book. It knew exactly what the book was about and its core ideas.
On this specific case, I cannot confirm whether or not OpenAI paid or has paid Donald Miller, the book's author, for using the content it created. What I can confirm is that there are several class action lawsuits and protests in the world, in which content creators complain that AI companies with GenAI solutions have used their content without permission, attribution or remuneration—a clear act of piracy.
If you want to go deeper on these cases, just Google "AI copyright lawsuits." Earlier this year, I published an article in Medium.com that report the case of the "empty" book titled "Don't Steal this Book." What happened was that, in the context of the London Book Fair, thousands of authors protested against AI firms using their content without consent, acknowledgement and payment, as well as against UK legislation proposing more lax copyright regulations.
Another example is a very recent article by The Washington Post covering the case of the New York Times and its lawsuit against OpenAI and Microsoft for using its content to train their models. If you have access to the WaPo, this is the article titled "Microsoft exec called AI the ‘largest theft of labor’ in history, court records show."
Yes, it is pretty cool that GenAI tools can write you a poem like if you were Shakespeare, or an article like if you were your favorite columnist, or even create a song mimicking your favorite artist, or any other things that involves the creations from artists and humans like you or me.
What is not cool is that this great convenience is at the expense of content creators who are pushed aside, while AI firms make subscription money from their innovative tools that cannot generate anything without the "data" from real creators. As a note and reference, the quarterly revenue of OpenAI is around US$6 billion dollars. Just saying.
Less meaningful human interactions?
Connecting people. It is one of the promises of most new technologies from the past decades. That is especially the case of instant messaging and social media. Fulfillment of this promise has not be proven.
Indeed, IMs, social media and other tools improve the quantity and velocity of communication and information exchange. However, it is not only clear the degree of quality and depth of those interactions.
If you are old enough to had lived part of your teen or young adult years without a smartphone, just compared how many people call you today on special events like your birthday, getting a new job or similar situations, versus 20 years ago when social media and short messages were still nascent.
In the context of the "dead Internet" and online communities, a study from Notre Dame University shows that reflexive bots that generated and share content increase user engagement. The downside is that—as the authors explain— this happens at the expense of human-to-human interactions. The mechanics of this involve humans interacting directly in online communities, bots jumping in into the conversation, and humans having fewer back-and-forth discussions.
What are some countries doing to deal with the implications of a dead internet in the times of AI?
Every regulatory decision should originate from a real and tangible local problem. The same applies to AI regulation. Although not guided by the "Dead Internet Theory", a few governments have started to pass legislation or take action on issues that somehow are related to the fact that AI is taking over online content and engagements.
Some of these actions mean to create protections. In other cases, they aim to create flexible arrangements about IP and copyright legislation, to accommodate for the disruption and new reality created by AI. Below, I share a few examples:
1. The European Union. As part of its AI Act's Article 50, which kicked in this past August, the EU requires developers and those publishing AI generated content to mark, label and disclose it, so end users become aware that it is synthetic content—either text, image, video or audio. Per se, this does not address the issues I have highlighted, but for sure brings transparency to AI-generated content. Other countries have taken or are considering similar regulatory requirements.
2. Japan. The country's Copyright Law provides a more flexible arrangement with some nuances. Its Article 30-4 allows the exploitation of copyrighted content for non-expressive purposes—including data analysis, testing and automated data processing, like training AI models. However, the same article prohibits these actions when there is unreasonable prejudice affecting the creator's interests. In addition, Article 47 opens the door for minor exploitation and activities that support computer data processing and online search results. Overall, Japan's rules are considered very flexible.
3. United States of America. There is no legal ban in the US for web scraping publicly available online content, unless this action bypasses password or login protections. Even Court rulings have decide that scraping public data on the web does not fall under the provisions of the Computer Fraud and Abuse Act, even if the action is against the site's Terms of Use of Service. At the same time, the Federal Trade Commission ruled on the use of consumer reviews and testimonials (16 CFR Part 465). It prohibits bots and AI tools deployed for deceptive commercial manipulation from writing, creating or distributing fake news and testimonials from users that do not exist or lack actual product experience.
4. Canada. In the context of its personal information protection legislation (PIPEDA), Canada allows collection, use and disclose of personal information only when there is consent and usage is related to activities considered as reasonable and appropriate. The law outlines exemptions when there is no consent, only if the information is publicly available and allowed by regulation. Consent exemptions do not include web scraping of email addresses or when hacking computer access to obtain information.
5. France. Recently the Alliance de la Presse d'Information Générale filed a complaint with the French Competition Authority, related to the impact of Google's AI Overview. According to a Reuters article, the group representing French newspapers and magazines have voiced that they have neither authorized Google nor received remuneration from the tech company for scrapping news and using their content on Google AI Overviews.
6. Australia. I close this list with a country that has set its intention to protect content creators. In a speech I recommend you to read, the Prime Minister was clear that there will be "no free pass for AI" in Australia. Here is a quote on copyright protection: "An artist’s creative endeavour is their work and their property. No company should use Australian books, music, art or news to build or train AI without the artist’s control. That includes the artist’s control of the price and value of their work. Anything less, is theft."
What are you going to do now that you know this about AI?
After researching on this topic and writing the lines above, my take is that instead of talking about a "dead Internet," we should be talking more about a new "synthetic Internet" or "AI-dominated Internet."
Beyond the labels you or I want to use, the question is: what are we going to do about this? Not as spectators, but as individuals being benefited and/or affected by this.
As users, we all like the benefits and convenience that disruptive technologies deliver. In my case, I was really satisfied when Uber and similar services emerged. Today, those days of dealing with poor quality cab services in the DC area are over. Similarly, many of us enjoy the benefits AI brings to the online world.
However, what happens when you are on the side of those affected by the new kids on the block, or as I said, the new artificial kinds on the block? Going back to the case of Uber, incumbent taxis were paying taxes, licensing fees and other costs from operating formally, when suddenly a more innovative competitor—Uber—comes in with a very similar service that operates under a big gray area, regulation-wise.
This is when governments need to take action, not necessarily by pulling the legal or regulatory trigger, but mainly by understanding issues and creating spaces for dialogue, to then make informed policy decisions.
However, that reaction does not always happen or is painfully slow for the affected stakeholders. When that happens, people start taking matters into their own hands. Sometimes this means a legal battle. Still, this is a path better suited for those with financial resources or when acting collectively (class action) versus individually. Other times, those affected adapt their business models to the "new reality." More often, those too small to fight or slow to adapt are left behind by the swifts of disruptive changes.
A deeper discussion requires reflecting on the purpose and value of the Internet, not only for AI firms but for all humanity. If that purpose and value is not affected, even if AI continues to take over online content and engagement, rules can be adapted to protect what needs to be protect. But if these changes shift the underlying value and purpose of the Internet towards a space that is more beneficial only for the few and less for the majority, then we might declare that, indeed, the Internet is dead.