"The study is based on data collected by the Internet Archive’s Wayback Machine from late 2022 (when ChatGPT kicked off the AI craze) to mid-2025. As of May 2025, the researchers found that 35.3% of all newly published websites on the internet were created with the assistance of AI, including 17.6% of websites that were completely AI-generated."
"The researchers found that AI-generated online content wasn’t as factually incorrect as expected. They also found that it cited its sources through external links, didn’t extinguish individual writing styles in favor of a generic voice, and wasn’t a long, winding block of text with little meaningful information, despite common belief."
I know Google AI gives the external sources links, so it's not a matter of just mindlessly "absorbing" its findings. You can see immediately in most cases whether the sources are even relevant much less credible, without even looking at them in detail. So far it's doing pretty well with "key highlights" type of summaries.