Binghamton University: More than half of new web articles are AI-generated
A study by Graphite, an AI content detection service, found that more than half of new articles published on the web are generated by artificial intelligence. Francesco Agniellini, a Digital and Data Studies lecturer at Binghamton University, examined the findings and published his analysis through The Conversation. The study covered more than 65,000 randomly selected articles of at least 100 words, spanning peer-reviewed content, promotional material, and general web publishing.
The concentration of AI-generated content is not evenly distributed. Machine-generated pieces cluster around practical, formulaic categories: lifestyle guides, product reviews, how-to articles, and SEO-optimized copy. These are also the categories that historically sustained the largest share of freelance writing work — which makes the displacement economically significant for writers who relied on that volume.
Agniellini does not frame this as the end of professional writing. His argument is that the shift creates a bifurcated landscape. Work that AI produces readily — high-volume, predictable, templated — becomes a commodity. Work that requires personal perspective, institutional knowledge, narrative originality, or accountability for accuracy does not become a commodity at the same rate. Authentic, distinctive voices may gain relative value in a web where generic content proliferates.
The analysis raises a separate concern about the effect of AI tools on writers who use them regularly. Research cited in the piece suggests that reliance on AI for drafting can narrow creative range over time, pulling writers toward uniform stylistic patterns. The risk is more acute for non-English speakers and has implications for cultural diversity in published writing.
Agniellini himself uses AI to refine drafts before editorial review, and he includes this disclosure as part of his argument — that human-AI collaboration is already a standard practice, not an exception. The more relevant question is not whether writers use AI, but whether the work that reaches readers carries traceable judgment, original perspective, and responsibility for what it claims.
Useful for editorial teams thinking through how to position human-written content in a market where AI-generated volume is growing, and for writers evaluating where to concentrate their effort.