Spam websites littered with fraudulent content are not new. However, with widespread use of AI, you can expect yourself to be more likely to land on an AI-generated spam website in the near future.
NewsGuard, a misinformation research organisation, released a report detailing the prevalence of AI-generated junk sites on the internet. As of mid-May, they identified a total of 125 websites “entirely or mostly generated by AI tools”. This is up 155% from 49 websites in early May. To date, NewsGuard says that there are 277 unreliable AI-generated junk websites.
But how exactly does one distinguish a legitimate, reputable website from a shady one?
If it looks suspicious, it probably is
If a certain website’s name sounds like it is an authority on a specific topic, but you have never heard of it, there is a chance that it is a red flag.
For example, NewsGuard’s research found websites with names like ‘Ireland Top News’, ‘Daily Time Update’, and ‘iBusiness Daily’ – these sites use generic, ‘news-sounding’ terms to appear as legitimate.
Reputable news sites like CNET have disclosed the use of AI to produce editorial content. If a suspicious website does not contain a disclosure notifying readers that it is using AI, this is also a red flag.
Unverified information
Drinking your own urine could cure cancer? The Egyptian pyramids were powered by electricity grids? Taylor Swift and Matty Healy are engaged and have a child?
If a website produces seemingly unbelievable information and passes it off as legitimate, it is probably a junk website. If reputable newswires such as the BBC, Bloomberg, Reuters, and Associated Press have yet to confirm it, it is probably a hoax. If an article has no author, or if the website does not specify its owner, these are also signs that a site may be AI-generated.
Keyword searches
One trick to find out if a website’s information is generated with AI is to do a site or keyword search.
According to NewsGuard, AI chatbots like ChatGPT tend to churn out generic disclaimers such as “as an AI language model…”, “I cannot complete this task,” or “I don’t have access to current events”. Phrases like “likely refers to”, “it’s important to remember”, and “in summary” are also indicators that a sentence or paragraph is written with ChatGPT.
To check if a website is presenting AI-generated information, simply perform the following:
- Head over to the Google search bar on your browser
- Type site: followed by the website address. For example, site:thechainsaw.com
- Hit the spacebar, and enter the above phrases with quotation marks. For example, site:thechainsaw.com “likely refers to”
- Hit ‘Search.’
AI-generated news is a thing now
… fortunately, NewsGuard has launched a tracker dedicated to documenting AI-generated websites in different languages. However, at the rate in which AI is developing, expect more AI-generated websites to flood the internet with misinformation. We hope the tracker can keep up.
