• Posted on 10 Sep 2026

2026 is turning out to be a big year for AI – and not just because of the amount of money pouring into the sector. Several research companies now confirm that the amount of AI-generated information on the internet has hit around 52 percent, which is more re-assuring than what was predicted by Europol back in 2022 when OpenAI brought ChatGPT to the world. Back then, the European law-enforcement group was predicting the amount of AI-generated material on the internet would hit a staggering 92 percent. All of which is perhaps telling us that, so far, we still like to consume information written by human beings.  

While 52 percent is a big number, there’s some solace to be taken from the latest report by the SEO company, Graphite, which came up with that number after it analysed a random sample of 65,000 English-language news articles published between January 2020 and May 2025. Any article in which more than 50 percent of content was AI-generated was deemed AI-generated. To be expected, there was a huge spike in AI-generated articles appearing just after Open AI’s ChatGPT launch, hitting a peak in November 2024, but then it slowed. Graphite repeated the study in early 2026, using three rather than 2 AI detectors, and found the number fell to 50 percent, and although that still sounds like a lot, it might be an under-estimation.  

There is a problem distinguishing between AI-generated and human-generated material when humans are working, often but not always, closely with AI to generate content. In fact, we use AI in so many different ways that it would be challenging to make safe predictions about the percentage of AI-generated vs human-generated content in any given piece of information appearing on the internet. As Stefano Soatta, professor of computer science at UCLA told Axios: “At this point, it’s a symbiosis more than a dichotomy.”   

Still, as Axios points out, in a follow up survey conducted by Graphite, the results show AI content is deprioritised by search engines and chatbot responses. In fact, Graphite found that only 14 percent of articles ranked in Google search were AI-generated; 86 percent were written by humans. And chatbots showed a similar trend with 82 percent of articles called up by ChatGPT and Perplexity having been written by humans.  

But the future comes fast in the world of AI. The 90 percent Europol figure would create a curious and no doubt frightening internet user experience filled with AI giveaways like repetitive familiar phrases, and more em-dashes and Oxford commas than are healthy. It might also pose serious threats to AI companies themselves.  

It could bring about model collapse where, as Axios points out, the quality of AI responses deteriorates because generative AI engines are feeding off AI data, rather than material produced by human beings. This model collapse has also been labelled Model Autophagy Disorder, which researchers at Rice and Stanford Universities identify as resulting from AI consuming its own product, and alternatively, ‘Habsburg AI’, a term which Monash University ARC fellow, Jathan Sadowski came up with to describe a future in which AI systems become inbred mutants. At least there’d be a movie in that! 

 

References

Europol Innovation Lab – Facing reality? Law enforcement and the challenge of deepfakes 
https://www.europol.europa.eu/cms/sites/default/files/documents/Europol_Innovation_Lab_Facing_Reality_Law_Enforcement_And_The_Challenge_Of_Deepfakes.pdf
 
Graphite - More Articles Are Now Created by AI Than Humans 
https://graphite.io/five-percent/research/more-articles-are-now-created-by-ai-than-humans 
 
Graphite - AI Now Writes as Many Online Articles as Humans 
https://graphite.io/five-percent/research/ai-now-writes-as-many-online-articles-as-humans-do 
 
Axios - Exclusive: AI writing hasn't overwhelmed the web yet 
https://www.axios.com/2025/10/14/ai-generated-writing-humans 
 
Graphite - How Does AI-Generated Content Perform in Search and Answer Engines? 
https://graphite.io/five-percent/research/ai-content-in-search-and-llms 
 
Axios - AI could choke on its own exhaust as it fills the web 
https://www.axios.com/2023/08/28/ai-content-flood-model-collapse 
 
ArXiv - Self-Consuming Generative Models Go MAD 
https://arxiv.org/pdf/2307.01850 
 
This Machine Kills (Paywall) 
https://www.patreon.com/thismachinekills 

Share

Author

Monica Attard

Co-Director, Centre For Media Transition, Faculty of Design and Society

News

My slop my way

Our latest Centre for Media Transition newsletter - My slop my way | Issue 15

News

On light duties

CMT's Michael Davis looks at how the proposed digital duty of care is still narrowly focused on harmful content. While addressing this more effectively is of...

News

Young people want to see themselves in the news

CMT's Harry Craigie looks at the latest report from the long-running News and Young Australians project, which found that news engagement amongst the young...

News

Big money, big winners

Centre for media Transition Newsletter - Funding journalism and building legal tech for social justice | Issue 14