Oversharing
When it comes to sharing news articles and videos on social media, more isn't always better
Sharing news articles and videos from media organizations large and small is a perennially popular use of social media platforms, but not all news sources are equivalent in their degree of commitment to fact, and not all users have the same sharing patterns. On platforms with open APIs such as Bluesky, account behavior can be studied at scale, and by taking advantage of MediaBiasFactCheck.com’s factual reporting ratings, can be correlated to the accuracy of news articles and videos being linked. It turns out that, at least on Bluesky in its present state, low quality news sources are more likely to be shared by high volume accounts, and by accounts less than a month old.
The website MediaBiasFactCheck.com maintains ratings for over 9000 news websites, scoring each source on political bias and factual accuracy, and offers an API where users can download the full set of ratings for free up to three times per month in JSON format. This can be used to augment Bluesky data with the MediaBiasFactCheck ratings for websites linked in Bluesky posts. For the sake of this experiment, all visible posts (excluding reposts) between April 10th and May 10th 2025 linking to any website rated by MediaBiasFactCheck were downloaded and annotated with their MediaBiasFactCheck ratings. The official feeds of rated news organizations were excluded from the dataset, leaving a total of 2669881 news links posted by 284927 Bluesky accounts.
Overall, Bluesky users linked articles with “High” MediaBiasFactCheck factual reporting ratings more frequently than any other category during the period studied, accounting for 62.9% of links, followed by the “Mostly Factual” category at 16.3%. Links of dubious accuracy (“Mixed”, “Low”, and “Very Low”) made up 16.9% of the volume. Newer accounts shared low quality sources at a higher rate, with 25.4% of links from accounts less than 30 days old having “Mixed”, “Low”, or “Very Low” factual reporting ratings.
The data shows a noticeable inverse correlation between factual accuracy of linked sources and an account’s daily posting volume. Sites with “Low” and “Very Low” MediaBiasFactCheck ratings were generally linked by accounts with substantially higher daily volume than sites with higher ratings. (The number of posts per day was estimated by downloading up to 1000 chronologically-ordered posts from each account and doing the math.) The cause of this phenomenon is unclear, although one plausible explanation is that people who share more articles and videos are less likely to have actually read or watched every article or video they share.
There are some obvious limitations to this experiment. First, while MediaBiasFactCheck maintains ratings for an impressively wide variety of news sites, their coverage of social media-based sources such as YouTube channels is a bit spottier, especially in cases where there is no official website. Secondly, the choice of Bluesky as the only platform to study was largely driven by the fact that larger platforms have significantly limited data access for third-party researchers; including X in the dataset would likely provide more comprehensive results, especially since Bluesky does not have a large population of users linking to right-wing news sites.