Nov. 24, 2025 /Mpelembe Media/ — The analysis of the growth and prevalence of AI-generated articles being published on the web clearly indicates that the quantity of articles produced by AI surpassed human-written content in November 2024, a significant trend spurred by the launch of ChatGPT in late 2022. However, the proportion of AI content has recently stabilised and suggests this might be due to AI articles often not performing well in major search engines like Google. The CommonCrawl dataset and the application of an AI detection algorithm has been known for its false positive and negative rates when using articles from before ChatGPT’s release and articles generated by GPT-4o, respectively.
To evaluate the false negative rate (the percentage of AI-generated articles classified as human-written), 6,009 articles were generated using OpenAI’s GPT-4o. These articles were generated using the OpenAI API with specific system and article prompts.
It is also worth noting that the study only evaluated the false negative rate on articles generated by GPT-4o. The AI detection algorithm might have lower accuracy when applied to articles generated by other models. SEO AI detection algorithm correctly classified 99.4% of these GPT-4o-generated articles as AI-generated, suggesting a 0.6% false negative rate for GPT-4o.
The shift of AI-generated articles surpassing human output on the web carries significant implications for both information accessibility and quality.
Implications for Information Accessibility
While the volume of AI content published on the web is now greater than human-written content, its accessibility to the end-user via major platforms appears constrained:
Low Visibility in Key Platforms: Despite the sheer prevalence of AI-generated articles on the web, these articles largely do not appear in Google and ChatGPT.
Poor Search Performance: The sources suggest that practitioners have found that AI-generated articles do not perform well in search. This poor performance is hypothesized to be the reason why the growth trend in AI-generated articles has plateaued since May 2024.
Limited User Viewership: It is suspected that AI-generated articles are not viewed in proportion by real users relative to their total volume published online.
Therefore, the primary implication for accessibility is a divergence: there is a massive volume of AI content being produced, but it often fails to reach the user through the primary information retrieval channels (search engines and large language models like ChatGPT).
Implications for Information Quality
The rapid improvement in AI quality challenges traditional assumptions about content quality and authorship
Improving Quality: The quality of AI content is rapidly improving.
Competitive Quality: In many cases, AI-generated content is considered as good or better than content written by humans. This statement references an MIT study.
Difficulty in Distinction: It is often hard for people to distinguish whether content was created by AI. This is supported by an Originality AI study.
Detection Challenges: The accuracy of AI detection algorithms is a subject of considerable disagreement, with some experts arguing that detecting AI is “impossible, or at best, highly inaccurate”. Furthermore, while one detection algorithm showed a low false negative rate (0.6%) on articles generated by GPT-4o, AI models continue to improve and may become more difficult to detect.:
Context: Why the Shift Occurred
The surge in AI content production, which began significantly after the launch of ChatGPT in November 2022, was driven by economic incentives:
Cost Efficiency: Using Large Language Models (LLMs) such as ChatGPT, Claude, and Gemini to publish content is viewed as a cost-effective alternative to paying humans hundreds of dollars to write content.
Traffic Strategy: Many companies explored this method to grow their traffic across channels including Google Search, social, and advertising.
It is also important to note that the study did not evaluate the prevalence of AI-generated/human-edited articles (AI-assisted content), which may be even more prevalent than purely AI-generated articles.
The situation regarding AI content proliferation is like filling a vast library with incredibly realistic replicas of books: while the total number of books (articles) in the library has exploded, most of the new, replicated books are being stored in remote stacks that the main public library catalogue (Google/ChatGPT) doesn’t index, meaning the user seeking information often encounters only the vetted originals.