A study by the SEO firm Graphite found that 52% of all new written content on the internet is now AI-generated. The research tracked a rapid increase in such articles following the November 2022 release of ChatGPT.
The analysis examined a sample of 65,000 English-language articles published between January 2020 and May 2025. To conduct the classification, the study utilized an AI detection tool named Surfer, which flagged articles as AI slop if they contained 50% or more machine-generated text. Before the launch of ChatGPT, AI-generated pieces accounted for approximately 10% of new content. This figure surged to over 40% by 2024, demonstrating a swift adoption of the technology for content production.
The proportion of AI-authored articles briefly exceeded that of human-written ones in late 2024 before settling. By May 2025, the split had stabilized near a 50-50 ratio, with AI accounting for the slight majority of new material. Despite this high volume, search engines like Google are reportedly filtering out low-quality AI “slop” from top search results. Consequently, only about 14% of Google’s highest-ranked content is identified as being AI-generated.
This rise is primarily driven by the cost-effective application of large language models, including ChatGPT, Claude, and Gemini, for search engine optimization (SEO) and to increase online traffic. The quality of AI content has also improved, in some cases matching or exceeding human-written text and making it more difficult to distinguish. Experts suggest the trend represents a symbiosis between human and AI content creation rather than a complete replacement of human writers.
Concerns have been raised regarding potential feedback loops. This issue involves AI models training on data that includes poor-quality or other AI-generated content, which could degrade the overall quality of internet content over time.
