I couldn’t resist coming up with this provocative title, when I recently stumbled upon an article in Forbes Australia [1] titled “Is AI quietly killing itself – and the Internet?” which is also quite provocative as a title. So, what is the topic here?
The rise of generative AI unfolded through a series of pivotal milestones. In November 2022, OpenAI introduced ChatGPT as a free research preview, sparking global curiosity and experimentation. By February 2023, the model had matured into a commercial product with the launch of a premium subscription. That same month, Microsoft embedded Copilot into its Office suite, integrating AI into everyday productivity tools and accelerating mainstream adoption. Then in July 2025, the debut of Perplexity Comet, an AI-native browser, marked a new frontier, where AI didn’t just assist but redefined how users navigate and interact with the Web.
So, it is only a few years now since artificial intelligence (AI) started to radically change the digital landscape and the Web in particular. This change has prompted an emergent paradox: AI may be undermining its own future. A recent study by researchers from Oxford and Cambridge, published in Nature [2] , reveals a phenomenon called model collapse, where generative AI systems degrade in quality when trained on content produced by other AIs.
Model collapse begins subtly; minority data and nuanced information are the first to vanish. Over time, the AI’s outputs become homogenized and eventually nonsensical. One experiment showed that after just nine cycles of self-training, an AI’s factual article on church steeples devolved into gibberish about jack-tailed rabbits.
The problem arises as AI-generated content increasingly saturates the Internet in all available media formats (text, speech, video…). According to another study [3], over 57% of web-based text has already been generated or translated by AI. If this trend continues, future AI models may be trained predominantly on synthetic data, leading to a feedback loop of diminishing accuracy, diversity, and truthfulness. A dystopic prophecy in January 2023 goes even beyond and forecasted that 90% of online content could be generated by AI by 2025. While I am writing this article in late 2025, I suppose we are not there yet.
The implications are profound. Without access to fresh, human-generated content, AI risks becoming a self-referential echo chamber. This threatens the reliability of future AI systems and the integrity of the Internet itself, as misinformation and hallucinations proliferate.
The solution? Sustained access to authentic, human-created data and global coordination to track content provenance. Without it, the digital future may be built on a crumbling foundation of synthetic illusions.
You can ask now what has this to do with the Internet per se? The Internet is only an information superhighway feeding the AI training modules with data and spreading the AI generated content to those who asked for it. A legitimate question.
An examination of how the Internet is financed reveals that major platforms, such as Google, Facebook, YouTube, and TikTok, rely almost entirely on advertising revenue. This income underwrites the infrastructure and the free content users consume daily. However, emerging analysis suggest that this model is under threat. One article, for instance, highlights that despite Google’s robust financial performance, its advertising foundation is quietly deteriorating due to AI-generated features like AI Overviews [4] . These summaries reduce user engagement with ads and links, jeopardizing the sustainability of ad-supported content. Ironically, Google’s own AI advancements may be accelerating the decline of its primary revenue stream.
Compounding this issue is a generational shift in career aspirations. Increasingly, young people say they want to become influencers; a role rooted in digital visibility, creativity, and autonomy. Yet this ambition carries a significant risk: the influencer economy is fundamentally dependent on advertising and social engagement metrics. If the ad model collapses, the influencer profession may fade before it fully matures.
I can speculate further, but I am confident that we will remember one of the dates mentioned in the beginning of this article as the date that caused a seismic shift on the Internet. Or maybe this date is yet to come in the not so distant future.
Transparency disclaimer
This article was written with the assistance of an LLM to enhance formulation and clarity of expression.
References
[1] Tor Constantino. (2024, September 3). Is AI quietly killing itself – and the Internet? Forbes Australia. https://www.forbes.com.au/news/innovation/is-ai-quietly-killing-itself-and-the-internet/
[2] Shumailov, I., Shumaylov, Z., Zhao, Y., Papernot, N., Anderson, R., & Gal, Y. (2024). AI models collapse when trained on recursively generated data. Nature, 631, Article 8022. https://doi.org/10.1038/s41586-024-07566-y
[3] Thompson, B., Dhaliwal, M. P., Frisch, P., Domhan, T., & Federico, M. (2024). A shocking amount of the web is machine translated: Insights from multi-way parallelism. arXiv. https://arxiv.org/abs/2401.05749
[4] Dans, E. (2025, July 25). The robot that ate Google’s profits: AI’s silent advertising apocalypse. Medium. https://medium.com/enrique-dans/the-robot-that-ate-googles-profits-ai-s-silent-advertising-apocalypse-9c0f3baa5d04