Low-quality news posts receive higher engagement but not more impressions on Twitter

One of the most influential findings in misinformation research is the observation that false news spreads farther than similar true news on Twitter. This finding has been widely interpreted as evidence that social media platforms preferentially expose users to misinformation. This work, however, was largely based on claims that were evaluated by fact-checkers - and true claims that were fact-checked are unrepresentative of true claims more broadly. Thus, it is unclear whether prior results will generalize to more representative content. Equally importantly, prior work examined how much content was shared, rather than how many users the content reached. Thus it also remains unclear whether misinformation - even if shared more - actually reaches more users online than accurate information. Here, we address these issues by analyzing over 2 million posts from a random sample of 23,885 Twitter users; as well as a separate sample of all news URLs shared on Twitter over a 9 day period. We find that posts containing links to lower quality news domains do in fact receive significantly more shares and likes than posts containing links to higher quality news domains. In contrast, however, we find no significant difference in the number of users who viewed posts containing links to lower versus higher quality news (and higher quality news receives a much larger amount of total engagement and views across posts). The disconnect between engagement and views may be explained by lower quality links being preferentially shared by users with insular networks and more bot followers.