Twitter and the spread of 'fake news': a new study by the CSA (FR)

posted on 22 June, 2021   (public)

A qualitative and quantitative study to better understand the mechanisms behind the dissemination of dis/misinformation on Twitter

Following the adoption of the French law on the fight against information manipulation in 2018, the French audiovisual media services regulator CSA released a two-tier study in November 2020 with the aim to better understand the mechanisms by which unreliable news are propagated on Twitter: a first analysis focusing on certain Twitter accounts labelled according to their reliability, and a second analysis based on a number of topical issues known to have given rise to the dissemination of disinformation or misinformation.

In June, the CSA published an English version of the report*.


The methodology of the analysis

The study aims at scrutinising the reality behind the spread of disinformation or misinformation, but does not monitor the compliance of Twitter with the legal requirements.

For this purpose, the CSA based its analysis on the combination of two kinds of data:

- The database of journalists specialising in fact-checking, and especially the "Decodex" by French daily newspaper 'Le Monde', identifying the degree of reliability of Twitter accounts and tweets;

- Data collected through the open API of Twitter.

The report also includes observations from Twitter on their policy (mechanisms to reduce visibility of certain content, transparency and access to data, their policy regarding Covid-19…).

The key outcomes of the analysis

The spread and visibility of disinformation:

  • Twitter accounts disseminating 'fake news' have less subscribers than 'reliable' accounts but a higher propension to see their content retweeted (10 to 20 times more). It is also worth noting that, according to the information published by the daily newspaper 'Le Monde', disinformation appears to attract particularly the 25 to 49-year-old age group and especially people from higher socio-professional categories. 
  • When it comes to disinformation, the CSA observed a high concentration of tweets in a short period of time, making the job of fact-checkers more difficult. As a result, "'genuine' news information does not displace fake news" and the verified information cannot achieve a majority position before the discussion fades away.


p.23 of the report.

For the purposes of this study, based on the 'Decodex', four categories of accounts were identified:

1: Parody accounts / 2: regularly disseminating unreliable news / 3: questionable reliability / 4: reliable accounts (professional journalists' accounts).


Content analysis from Twitter accounts disseminating disinformation/misinformation:

  •  Most of the content of Twitter accounts disseminating disinformation is related to politics, immigration, health, religion and terrorism (more than half of the corpus studied), while fact-checkers accounts focus mainly on media news and content related to media literacy.
  • Disinformation content is mainly used as a means to criticise the authorities or to express a sense of panic. The use of non-text content (images, videos, links, etc.) seems to characterise a strategy for increasing the legitimacy of published content.

A lack of evidence for 'echo chambers':

  • Less than 20% of subscribers to accounts spreading disinformation do not follow any reliable account and the “echo chamber” phenomenon remains low.


Even though the volume of tweets does not reflect the real audience for verified information which remains more important (more subscribers to reliable accounts), for some topics, however, 'fake news' remains more visible than verified information owing to the virality of its content.


Source: CSA France

* Study based on tweets in French language only.