To tweet or not to tweet: three-year experiment finds no citation bump from Twitter promotion

In a controlled experiment spanning three years and involving 11 scientists from a range of life science disciplines, a new study published in PLOS ONE demonstrates that sharing a paper on Twitter did not increase citations.

It has long been asserted that papers shared on social media platforms, such as X (formerly called Twitter), receive a higher number of citations. However, a team of leading science communicators, led by University of Washington School of Aquatic and Fishery Sciences Professor Trevor Branch, with a combined total of 223,000 followers at the end of the experiment, wanted to explore whether increased citations was actually caused by tweeting about them, or was a correlation due to other factors such as the quality of the paper and the importance of the information in the paper. In other words, some papers may simply be unusually good or important, resulting in both more shares on social media and more citations. 

Lead author Branch said “A group of us (on Twitter, of course) agreed that a controlled experiment was needed to test this idea, and we had a ton of fun doing the experiment and writing up the paper. But it definitely tested our patience waiting for the results.”

An example of a tweet sent out about a paper by Dr. Dani Rabaiotti, one of the scientists involved in the experiment.

The team conducted their experiment every month over more than a year starting in December 2018: randomly selecting five articles published in the same month and journal and tweeting about only one while retaining the others as controls, then waiting three more years for citations to accumulate. By recording information such as number of tweets, Altmetric scores (a measure of online attention from social media and news articles), and citation counts, the results demonstrate that although downloads and online attention both increased substantially for the tweeted articles, there was no detectable difference in the number of citations after three years. The authors note that there can be a long lag before additional citations manifest, perhaps longer than the three year study.

Number of downloads of the tweeted papers (blue) and untweeted control papers (red) showing how downloads increased substantially in the days immediately after tweeting.

Social media remains an important way to bridge the gap between scientific research and educating the public. It allows scientists across the globe to not only share their work with other scientists working in their areas of expertise, but also to reach a diverse audience from the non-scientific community. From the public to policy-makers, communicating about science in an engaging, understandable way is useful to demonstrate the importance of conducting research across a number of topics that have global impact and interest. However, as the team found out, doing so does not bump up citation counts, and this should be considered when thinking about the role of science communication. Is the goal to share work to broadly educate or to increase the profile of scientific papers?

“It’s important for scientists and technical experts to break out of the ivory tower and engage with the public, and social media tools are a powerful way to do that. Public science engagement results in a lot of good things for science and for society,” David Shiffman, research associate at Arizona State University shared. “However, tweeting about your published papers does not result in more citations for those papers, despite many people claiming otherwise. It turns out that the relationship between citations and social media shares is not causative – some papers are shared more on social media and cited more for other reasons, which a colleague calls the “good papers are good” hypothesis. If you want to talk about your research using social media, there are lots of reasons why I think you should do that, but don’t think it’ll get you more citations for your research.”

The scientists and science communicators involved in the study spanned three continents from the following institutions: 

Back to Top