Despite 46% of people lacking confidence in AI, its use continues to grow. This democratization has consequences on how we use the Internet. What impact has AI had on our way of searching for information, and how are tech giants staying competitive ?
The AI content invasion : how generated content is transforming the Internet ?
Today, when you go on the web to search for any information, you will see even more AI-generated content. For instance, on Amazon, it’s possible to buy a book, that explains what kind of mushrooms are edible. However, these books are entirely generated by AI and can be dangerous for people because they encourage them to eat deadly mushrooms. How can we explain that ? For different companies, it represents an advantage to produce this type of content because it’s cheap, and they can earn a lot of money in a short period of time. Most of the time, people who are producing these fake content are based in Africa or Asia and do this for a living and learn through online tutorials. So, it represents an above-average salary for them.
Another reason explains this spate of AI-generated content. Large language models (LLMs) like ChatGPT or Gemini are friendly and familiar. They generate content quickly, easily, and concisely, with personalized responses tailored to each user. Trained to use natural language, they represent an accessible technology for ordinary people. This contrasts sharply with Google, where users must formulate keyword-based queries and synthesize results from multiple sources. Speaking of Google, another point deserves attention. The platform has become cluttered with pop-ups about cookies, entertainment, and advertisements, making it increasingly unpleasant to use. ChatGPT, by comparison, offers a simpler, more streamlined experience.
In addition, AI can perhaps replace the traditional social networks as we know them, like Instagram and TikTok. Indeed, Meta has announced the launch of its new social network, Sora, in which all videos are generated by AI. This trend explains our relationship to social networks. Today, we use them as entertainment content and not only as a way to keep in contact with our family and friends. The purpose has changed, and requires the tech giants to adapt their strategies.
The vicious circle of AI content : quality degradation and democratic concerns
However, this trend to use even more AI content for our relationship with the web is problematic. In recent years, a new expression has emerged : we talk about « slop ». This illustrates the abundance of AI content that we can see on the Internet and which are sometimes useless or even misleading. It is possible to find this phenomenon in several applications like Instagram or Spotify, in which we can find music entirely produced by AI. Another example is that of « Shrimp Jesus » that we saw on the web some time ago. This event, which is funny at first glance, illustrates the poor relevance of certain content. Overall, it contributes to harming the quality of the web. We also speak about the « shittification of the Internet ».
This problem worries some professionals. For instance, a researcher at the CNRS center told the newspaper Le Monde that this change of habit in our way of searching for information can be problematic in terms of democracy. How can we explain that ? The operation of LLMs is based on statistics. In other words, ChatGPT calculates the probability of a word coming after another one. As a consequence, AI is searching for a standardized response and can also make other content, which can be useful and relevant, invisible.
Furthermore, the use of new technology makes the formation of a gap between us and the source more likely. Indeed, we don’t read different pages of the Internet, and we are satisfied with AI’s responses. Although humans are naturally lazy, it contributes to reinforcing this personality trait. This makes debate, exchanges and the assimilation of different sources of knowledge more difficult.
As a consequence of all this, we can assume that people are going to produce even less content on the traditional Internet, so there will be even more content produced by AI technology. Considering that AI needs content to train its technology, we would say that it will be mainly trained by itself, on what it has produced. In other words, it is just a vicious circle in which the value of AI loses all meaning and moves towards content that is completely devoid of soul and character. Another problem is that it gives a lot of power to AI companies and confirms once again the problem in terms of democracy.
Google’s AI overviews : solution or final break with the open Internet ?
To bring a response to this phenomenon, Google, leader in the field of AI, but also as a search engine, has decided to take the turn so as not to be left behind. It wants, through a technology named « AI Overviews » to use LLMs to give us results for our research. The functioning is simple : we ask a question to Google, then it suggests a summary of the different pages that we can traditionally find, without clicking on a link. Although this technology currently only exists in the USA, it could mark a definitive turning point in our use of the Internet and represent a break with the open Internet originally envisioned by Tim Berners-Lee. Indeed, this model moves away from an Internet without discrimination, blocking, throttling, or prioritization.
https://www.arte.tv/fr/videos/122187-000-A/l-ia-va-t-elle-tuer-internet/
https://www.forbes.fr/technologie/ia-les-statistiques-marquantes-a-retenir-en-2025/
https://www.blogdumoderateur.com/ai-overviews-comment-ia-generative-changer-recherche-google-seo/
