Viral Tweets, Fake News and Social Bots in Post-Factual Politics: Post-Truth PR and the French Presidential Elections 2017

International audience•PurposeIn the wake of Brexit and the 2016 US Presidential Elections, “post-factual” society has been heralded as a new era of political communications, where the digital public sphere plays a central role, in spreading “viral” contents and “fake news”, with the help of automated accounts or “social bots”. This paper seeks to define these terms and the methods by which the phenomena they commonly designate might be studied, in order to characterise the dynamics of political deliberation during the 2017 French Presidential Elections on Twitter, the online platform most commonly used for political communication in France. It thus aims to better understand the mechanisms by which information becomes popular and circulates rapidly on the network, notably the degree to which “fake” information is spread virally, and the role which bots may play in this, and to help public relations and political communication professionals to better take these into account in their strategies.•Design/methodology/approach The paper stems from an interdisciplinary research project involving computer scientists and communication scientists. It is based on a corpus of around 50 million Tweets constituted over 7 weeks during the French Presidential Election campaign in 2017. Focusing on a sub-corpus of just over 10 million tweets sent during the final two weeks of the election (between the first and second rounds of voting), the most popular retweets were selected (n=197), and subjected to qualitative analysis (manual coding) in order to characterise them in terms of their contents. In parallel, the Twitter accounts which has retweeted the most often (at least 100 times) the 1000 most common retweets were selected as the likely “influencers” of the sub-corpus (n=1077). The software tool specialised in bot detection, Botometer, developed by the University of Indiana, but also two types of algorithmic cluster analysis, were used to calculate the probability of automation of these accounts and to model the relationships between them. On the basis of this, different profiles of potential bots were identified. •Findings Results show that relatively little “fake” information featured among the most popular retweets during this period, and indeed that the expression of opinions, humour and irony, as well as denunciations of fakes and scandals, appear to be more likely to spread “virally”. There is clear evidence of at least some social bots, but detection techniques appear to be currently reaching their limits, at least in part because of increasingly sophisticated algorithms designed to escape detection, the use of automation tools to automate certain tweets only on their accounts by some individuals, and forms of human behaviour (e.g. bursts of rapid retweets) which resemble behaviour traditionally seen as that of bots. For these reasons, it would appear that bots will become relatively undetectable in the near future.•Research limitations/implications Limits to the study include the choice of sample and subjective bias introduced by the methods used (qualitative approach, cluster analysis), as well as the increasing difficulty of detecting social bots. The fact that “fake news” was not highly represented in the most retweeted messages is significant, but does not mean that “fake” information does not circulate on Twitter more widely. Implications for communications professionals seeking to optimise digital campaigning strategies on Twitter include insights into the types of contents likely to be spread virally, if other conditions (e.g. network structure and visibility) are fulfilled. •Originality/value The study is innovative in the choice to study virality through the prism of message contents, and also in the methodology of bot detection, which goes some way to address certain weaknesses of the current state-of-the-art solution. It is also original in seeking to bring together questions of virality, fakes and bots, questions which have no common definitions and generally treated separately, in order to propose a holistic reading of online political deliberation dynamics in post-factual society, through the particular example of the French Presidential Elections