“Dead Internet” is now a reality: robots dominate the web and are used to manipulate responses from Google and ChatGPT

The idea seemed exaggerated. However, two recent news They once again gave strength to that hypothesis.

On the one hand, Cloudflareone of the world’s leading internet infrastructure companies, assured that Bots now generate more web traffic than humans for the first time in history.

On the other hand, moderators of Reddit reported that companies are using online communities to influence responses that produce artificial intelligence tools such as ChatGPT and search systems powered by AI by Google.

The combination of both phenomena fuels a growing concern among experts, moderators and users: an Internet where machines not only produce more and more content, but also begin to influence what other machines will respond in the future.

The Dead Internet Theoryor dead internet theory, began to become popular in 2021. His proposition was simple: much of the activity we see every day on the web no longer comes from people, but from automated systems.

At that time, the hypothesis was dismissed by many as an exaggeration. But the advance of generative artificial intelligence changed the scenario.

Models like ChatGPT made it possible to produce texts, images and conversations on an unprecedented scale. At the same time, the number of automated accounts capable of interacting in an increasingly human-like way grew.

Even Sam AltmanCEO of OpenAI, recently acknowledged that he may have underestimated the phenomenon.

“I never took the dead internet theory very seriously, but it seems like there are now many accounts managed by language models”, he wrote on the social network X.

The observation is striking because it comes from one of the main drivers of modern artificial intelligence.

Now, a new data adds fuel to that discussion.

Matthew PrinceCEO and co-founder of Cloudflare, revealed that traffic generated by bots has already surpassed human traffic. According to the figures released by the company, currently the 57.5% of web requests come from automated systemscompared to 42.5% generated by people.

Prince acknowledged that the change happened sooner than he expected. Some time ago I had estimated that this turning point would only arrive in 2027.

The company clarifies that these are not just traditional bots used to index web pages or perform automatic tasks. Much of the growth comes from a new generation of agents powered by artificial intelligence capable of browsing sites, collecting information, comparing products, checking prices and execute complex tasks on behalf of the users.

Although humans continue to spend the most time consuming content, watching videos or using applications, the number of automatic requests generated by these systems already exceeds human activity measured in web traffic.

As bots gain ground on the internet, another phenomenon has begun to worry some of the largest communities on the web.

Moderators of the subreddit dedicated to biohacking reported that companies linked to the sale of peptides and hormonal therapies are using the forum to influence future responses generated by artificial intelligence.

As they explained, the problem is not traditional spam. The posts they detect often present themselves as legitimate questions or seemingly genuine discussions, designed to generate engagement and comments.

The strategy would take advantage of an increasingly relevant fact: platforms like Reddit are frequently used as a source of information by artificial intelligence tools and search systems powered by AI.

The moderators maintain that Some companies are looking to install certain themes, products or brands within popular conversations to increase the likelihood that that content will later appear in responses generated by systems like ChatGPT.

One of the forum administrators assured that some companies even try to understand What types of questions do language models usually prioritize and then replicate those formats? within the community.

The situation becomes especially complex because many of these accounts do not show obvious signs of manipulation. According to the moderators, They tend to be profiles with apparently normal activity and background similar to that of any user.

The difficulty in distinguishing authentic content from coordinated posts led the moderation team to restrict discussions about peptides and hormone therapiesconcentrating them in specific spaces to facilitate control.

Behind that decision appears a broader concern.

“There is a sadness seeing how this place on the internet that was so human it is eroding and getting stuck with artificial content powered by AI,” said one of the moderators.

The reflection summarizes one of the central fears behind the dead internet theory.

If an increasing portion of the web is produced by automated systems, and new artificial intelligences use that same web as a source to learn and answer questions, The risk is that the internet ends up feeding on itself.

In this scenario, machines do not only generate content. Also they condition the information that other machines will consume later.

The dead internet theory was born as speculation about a future dominated by algorithms. The latest data does not prove that this process has been completed. But they do show a reality difficult to ignore: Bots are already the majority in web traffic and more and more actors are trying to influence the responses of the artificial intelligences that millions of people use every day.

The question is no longer whether machines participate in the Internet. The discussion begins to be How much of what we see, read and believe online is still truly human.