According to a study by NewsGuard, there are a growing number of news sites that appear to be almost entirely written by artificial intelligence software.
OpenAI through ChatGPT has made it possible for everyone around the world to learn about Generative AI. Although its potential was immediately appreciated, questions were also raised about its use. There have been discussions about the possibility that it could generate fake news, or that it could reduce the ability of students to develop a personal style or lead to a reduction in jobs. As always, when you have a tool in your hands, it all depends on how you decide to use it.
Now a new study by NewsGuard has identified 49 industry-focused news sites that use AI to generate entire articles written in several languages, including English, Chinese, French, Portuguese, Thai, Tagalog, and Czech. Many of these sites appear to be real content farms run by anonymous sources that produce articles for the sole purpose of collecting advertising revenue. This practice has spread throughout this year as AI-powered tools have become widely used by Internet users.
The use of AI chatbots has also raised serious questions about how the technology can generate content, as tools like ChatGPT sometimes invent information out of thin air or report incorrect information. Human presence for content control seems to be crucial. Just think of the fact that Wikipedia itself is planning to use AI to create thousands of articles.
OpenAI said it uses a combination of human reviewers and automated systems to identify and combat the misuse of its model, including issuing warnings or, in severe cases, banning users. Google has emphasized that it rates content based on its existing editing rules and that using automation to generate content in order to manipulate search rankings violates the company’s spam policy.
In the case of the 49 sites identified by NewsGuard, only two of them admitted to using artificial intelligence but in most cases, it was difficult even to be able to contact the administrators to ask questions. What has been found is that in most cases the articles have mostly content summarized or rewritten from other sources and have errors, repetitive sentences, and clearly AI-generated sentences, such as “I am unable to produce 1500 words … However, I can give you a summary of the article”. This also highlights that there is no control by an editor before publication, increasing the risk of spreading incorrect news.
This study raises concerns about the use of AI in some contexts and the possibility that inappropriate use could contribute to misinformation.