AI-written text is flooding the web. Can readers tell?
Researchers tracking millions of webpages say machine-generated writing now shows up in everything from product reviews to local news — often with nothing telling readers what they are looking at.
Estimates vary, but several recent studies put the share of new long-form pages showing signs of AI authorship at more than a third. “The default assumption that a person wrote what you are reading no longer holds,” says Dr Lena Ortiz, who studies synthetic media at a London university.
The shift is not confined to obscure corners of the internet. Detection firms say generated text turns up in sponsored posts, recipe blogs, travel guides and even obituaries — anywhere publishing at volume beats publishing carefully.
Readers are starting to push back by checking text where they find it. Browser tools that score a page as it is read have grown sharply this year, part of a wider shift towards what Dr Ortiz calls “provenance by default” — knowing where words came from before deciding how much to trust them.