YouTube is still struggling to figure out how to keep from recommending divisive content like conspiracy theories.
On Monday afternoon, YouTube users on Twitter highlighted problems with the site’s autocomplete feature, which automatically suggests search queries after a user starts typing a few letters or a word, as well as its recommendation engine:
The thread was spawned by the New York Times’ recent editorial exploring YouTube’s role as “one of the most powerful radicalizing instruments of the 21st century.”
The video site, owned by Alphabet‘s Google, has been the subject of several recent investigations showing how it highlights extreme content, like conspiracy theories or hyper-partisan points of view, over more measured videos.
For example, YouTube and Google’s autocomplete boxes deliver starkly different answers, with the video site recommending controversial or fake points of view on YouTube.
YouTube results are on the left while Google results are on the right. Both searches were done in Google’s “incognito mode” to prevent the results from being affected by prior search history.
When The Wall Street Journal asked YouTube about this issue earlier this year, a Google spokesperson said other parts of the company were working with the YouTube team to “share learnings” about how to surface more authoritative content.
Google says that its autocomplete feature takes into account factors like “popularity” and “freshness,” though a representative didn’t respond to questions about how YouTube’s autocomplete feature differs.
The autocomplete and recommendation options are particularly jarring because Google’s Chromebook laptops, which include YouTube’s app, have become incredibly popular for students, making up more than half of all new laptop shipments for K-12 students in 2016, according to Futuresource Consulting.