Russian media interference in EU referendum worth up to £4 million
- Report submitted to the Culture, Media and Sport select committee by digital agency 89up lays bare extent of Russian media interference
- Kremlin-backed outlets RT (Russia Today) and Sputnik had more reach on Twitter for anti-EU content than either Vote Leave or Leave.EU
- Call for parliament to investigate role of UK-based Russian propaganda outlets
- Russian bots reach during EU referendum was significant - equivalent to 29% of total reach of both Vote Leave and Leave.EU Twitter activity
2. The true scale of Russian
media interference in
British democracy has not
been exposed until now…
3. Key findings
● Russian state media in the UK had almost the same social media impact as
the two main official Leave campaigns.
● Russian state media had a significantly larger impact on Twitter than the
two official Leave campaigns combined, suggesting the potential work of
bots.
● Russian state media interference in the EU referendum campaign had a
value estimated at between £1,455,500 - £4,140,000.
● Russian bots delivered 10,000,000+ potential Twitter impressions during
the EU referendum, with a reach nearly ⅓ of the Leave.EU account.
4. 5 - 44: Section 1: the role of the Russian state
media in the UK during the EU referendum
45 - 65: Section 2: the role of Russian bots
during the EU referendum
66: Appendix A: Terminology and Notes
5. INTRO
Who is 89up?
89up is a communications agency with a specialist social media analytics team led by Josh Feldberg.
Mike Harris provided insights into Russian propaganda methods.
How did we do this work?
Data was sourced from the Twitter Search API, Buzzsumo, the Facebook API and other scraping
methods. For equivalent media spend information we used data from Kantar.
Who paid for this work?
No one. 89up undertook this work because as a team we are concerned about the role of autocratic
regimes in shaping public opinion in democracies.
6. Section 1: the role of the
Russian state media on UK
citizens during the EU
referendum
There has been a lot of speculation
about the role of the Russian
government in the EU referendum in
2016. To date interest has focused on
the impact of Russian bots.
This report outlines the evidence for
Russian interference that was hiding
in plain view during the EU
referendum.
7. The real story of
Russian interference
in the EU
referendum was
hiding in plain sight…
Russian state media
propaganda through
Facebook and Twitter
led to significant
online traction
8. 261
Kremlin-aligned media published a significant number of unique
articles about the EU referendum, or anti-EU articles from January
2016 until the day of the referendum. Our researchers analysed the
most shared of these articles, and identified 261 articles with a clear
anti-EU bias to the reporting. The two main media outlets were RT and
Sputnik, with video produced by Ruptly.
9. The total value of Russian media for the Leave campaign in
the six months before the EU referendum was £1,353,000
The PR value for the Leave campaign of the 261 heavily pro-Leave articles
published by RT and Sputnik is estimated at nearly a million and a half pounds
based on figures from a leading media monitoring tool (6). This excludes the
significant social media value of these news articles.
The figure is likely to be much higher. If
articles with more neutral headlines -
that potentially embedded negative content
(see slide 21) this figure rises to £4,000,000.
This also excludes the value of the
TV broadcast by RT on Freeview.
10. There was a
heavy Leave
bias to the most
shared RT and
Sputnik articles
Of the 200 most
shared RT and
Sputnik articles
referencing the EU
referendum or Brexit
we analysed, there
was a significant bias
towards positive
coverage of the
Leave campaign.
Analysis of article bias
11. The bias is most
stark when
compared to other
media outlets
Whereas PBS covered
announcements of the
Leave and Remain
equally, with the majority
of their coverage clearly
neutral; RT’s coverage
was biased towards
highlighting the activities
of the Leave campaign.
Analysis of potential media bias
12. No other state controlled
media outlet came close to
social exposure of Russian
media
Compared to other major foreign
state-funded media outlets, the
Russian media outlets led to
significantly larger social media
exposure. Here we include all shares
for France24 and PBS (1), but we
narrow RT and Sputnik to only
pro-Leave articles - still the reach is
markedly greater.
Social engagement by foreign
state-funded media platform
13. Russian MediaVote Leave Leave.EU
Including only
objectively anti-EU
articles, Russian
state media articles
led to nearly as
many engagements
as the official Vote
Leave website.
When neutral articles on
RT/Sputnik are filtered out,
it is still the case that the
Russian media had nearly
as much social
engagement with anti-EU
articles as the official Vote
Leave campaign.
263,000
287,000
1,700
14. Russian Media on
Facebook
Vote Leave
on Facebook
Facebook:
Vote Leave spent $1.7
million on Facebook
paid impressions. The
Russian media
delivered an estimate
of 8,740,000 Facebook
impressions just with
anti-EU content.
The equivalent value (using
Vote Leave’s CPM) of
estimated Russian media
Facebook impressions would
be $102,000.
146,000,000
8,740,000
15. Russian Media Vote Leave Leave.EU
Twitter:
Russian state
media articles won
the Twitter war -
with significantly
more potential
Twitter impressions
than any content
from either anti-EU
campaign website.
We estimate the value of
Russian media’s potential
impressions on Twitter to
be between $47,000 -
$100,000 (8).
134,000,000
33,000,000
11,000,000
16. Why are RT and Sputnik different from other
media outlets?
News by mainstream UK publications - from newspapers to online magazines
to the majority of our TV - is paid for commercially either by advertising,
subscriptions or sales. The BBC - Ofcom-regulated and founded by Royal
Charter - has a legal duty to provide impartial news.
Online articles from RT or Sputnik UK are not regulated, either through
self-regulation (IPSO) or statutory regulation (Ofcom, IMPRESS). Some online
platforms are not subject to regulation or standards codes - but the clear
difference is that they do not receive a significant state subsidy from the
Russian government. Sputnik UK has an operating budget from the Russian
government of £1.8m a year. There is little evidence of any commercial
revenue except low revenue from Google adverts on the websites.
17. Why are RT and Sputnik different
from other media outlets?
RT (formerly Russia Today) has an international operating budget
of around £250 million per annum (7). Its expansion in the UK prior
to the EU referendum was seen as a way of increasing the
Kremlin’s influence. As Richard Sandbrook, director of the centre
of journalism at Cardiff University puts it: “It’s not a commercial
proposition, therefore the main purpose must be to gain
influence. It’s about soft power for the Kremlin”.
The Russian government is paying for biased news to be
published in the UK to influence British voters.
18. RT and Sputnik delivered
extreme anti-EU content
that was seeded across
the UK
The total number of impressions delivered by Russian
media won’t ever been known unless Facebook and
Twitter give the Culture, Media and Sport select
committee data.
For instance, an RT piece playing up French
Euroscepticism on the eve of the EU referendum was
shared 268 times directly from the RT Facebook group.
We will look at this single piece of content to show the
power of the Russian media.
19. Once posted, this single RT
post was seeded by other
Facebook groups from across
the political spectrum,
reaching between 11,800 (4)
and 39,310 people.
RT UK Facebook page
292k likes
325k followers
“Page of Info” Facebook
page
1.5k likes
1.5k followers
“UKIP Hemel Hempstead”
Facebook page
3.6k likes
3.3k followers
“Traditional Britain Group”
39k likes, 41k followers
“Huddersfield TUC”
22k likes, 21k followers
20. 167,144 estimated
Facebook impressions
Plus: 848,500
potential Twitter
impressions
To give an insight into the potential impressions per post we
analysed around 5% of the shares (252 of 4,934 shares) of one
of top 15 ranked RT posts on Brexit during this period. These
252 shares came from RT UK posting the content on their
owned Facebook channel. From this single post we estimate
the potential Facebook impressions delivered were a
minimum of 11,800 potential impressions from shares to the
major groups shown on the previous page, plus a further 248
shares by individuals with estimated reach of 8,382 potential
impressions (5).
Beyond this, there were a further 4,682 shares of which 4,348
were on Facebook and 323 on Twitter.
We estimate these Facebook shares, if individuals would
deliver 146,962 impressions.
A data analysis of the Twitter shares shows this article on
Twitter led to 848,500 potential impressions.
21. Even ‘impartial’ news
reporting embedded
anti-EU content
Even more “impartial” news reportage on RT and Sputnik
would lead a reasonable reader to a more Eurosceptic position
through the use of visual links that led to Eurosceptic content.
Often, embedded in the heart of neutral reportage would be
links to more overt Eurosceptic news stories.
Here we present one article which was a fairly impartial
assessment of the latest polling data during the EU
referendum. However, inside the article were embedded a
series of leading tweets from other RT articles.
22. Although the content of the piece was relatively neutral,
embedded in the piece were three tweets with leading titles.
23. Facebook Engagements +
Twitter / Instagram /
LinkedIn / Pinterest shares
Facebook
Engagements
Estimated potential
Facebook
Impressions
Potential Twitter
Impressions
Vote Leave 319,901 287,491 16,436,000 33,000,000
RT 215,337 191,139 7,349,568 120,000,000
Sputnik News 47,813 40,465 1,390,333 14,000,000
Leave.eu 1,970 1,729 489,190 11,000,000
* See note (5). estimated Facebook impressions based on number of shares on Facebook and average number of friends
per user, 338
Social media reach of Russian media anti-EU articles vs. Leave campaign
24. Russian media in the UK has a developed
social following
292k likes
325k followers
66K Twitter followers
54k YouTube subscribers
278k Instagram followers
(global channel)
2.7m likes
3m followers
55.4k Twitter followers
45k YouTube subscribers
17.8k Instagram
followers
(all global channels)
9k likes
9k followers
7.5k Twitter followers
24k YouTube subscribers
(global channel)
30.7k Instagram followers
230k likes
244k followers
52.5K Twitter followers
304k YouTube
subscribers
1k Instagram followers
(all global channels)
26. Key questions for Facebook and Twitter
● How much did RT, Sputnik and Ruptly spend on advertising on your
platforms in the six months before the referendum in 2016?
● How much have these media platforms spent to build their social
followings?
● Sputnik has no active Facebook page, but has a significant number of
Facebook shares for anti-EU content, does Sputnik have an active
Facebook advertising account?
● Will Facebook and Twitter check the dissemination of content from these
sites to check they are not using bots to push their content?
27. Key questions for Facebook and Twitter
● Did either RT, Sputnik or Ruptly use ‘dark posts’ on either Facebook or
Twitter to push their content during the EU referendum, or have they used
‘dark posts’ to build their extensive social media following? (11)
● What processes do Facebook or Twitter have in place when accepting
advertising from media outlets or state owned corporations from
autocratic or authoritarian countries? Noting that Twitter no longer takes
advertising from either RT or Sputnik.
● Did any representatives of Facebook or Twitter pro-actively engage with
RT or Sputnik to sell inventory, products or services on the two platforms
in the period before 23 June 2016?
31. Overview
1,730 is the total number of Twitter
posts on Brexit by Russian media from:
- Russian Today
- Ruptly
- Sputnik News
These tweets had a combined reach of
970 million total potential impressions
32. (Day of Brexit) Russian media tweeting Brexit updates,
with tweets of both anti-Brexit and pro-Brexit sentiment.
On the day of the EU referendum, Russian media significantly increased their push on Twitter on
referendum stories
35. Our analysis of bias
Our researchers assessed the most shared articles about the EU or the EU referendum debate
between 1 January and the close of polls on 23 June 2016. We divided the articles into three
categories based on the criterion: What would a reasonable reader infer the bias of this article was? A
number of the RT/Sputnik articles were straight reporting of the facts, or had quotes from both sides
of the debate. We classified these as ‘neutral’, e.g. there was no objective bias. For articles or
comment pieces that emphasised positive aspects of the European Union, or of the Remain
campaign, we classified these articles as ‘Remain’. No social share data for articles that were Remain
or Neutral is included in our analysis of anti-EU articles. Articles that a reasonable person would infer
had a bias towards the Leave campaign, or were overtly negative about the European Union, we
classified as ‘Leave’ - these are the articles that we analysed in more detail and that the social data
relates to.
36. Russia Today
We analysed 98 pro-Brexit or anti-EU articles from RT.com between 1 January 2016 and 23 June 2016.
These were shared or engaged with a total of 270,781 times. Breakdown of shares by platform:
- Facebook: 215,337 engagements
- Twitter: 21,816 shares
- LinkedIn: 466 shares
- Pinterest: 162 shares
Twitter:
- Shared by over 10,500 unique users from the RT Twitter channel.
- 74% of these users shared just one of the articles. 16% shared more than one
37. Sputnik News
Between 1 January 2016 and 23 June 2016 Sputnik News published 168 pro-Brexit or anti-EU articles.
These were shared or engaged with a total of 47,813 times. Breakdown of shares of this anti-EU
content by platform:
- Facebook: 40,465 engagements
- Twitter: 6,861 shares
- LinkedIn 93 shares
- Pinterest 155 shares
Twitter:
- Tweets from the Sputnik Twitter account about Brexit were shared by 3,389 unique users.
- 81% of these users shared just one of the articles. 19% shared more than one
- There's a 7% overlap between people who retweet Sputnik and Russia Today
43. Key findings
● Russian bots did play a role in the EU
referendum.
Russian bots identified by the US
Congress and Buzzfeed delivered over
10,000,000+ potential Twitter impressions
during the EU referendum. This is
equivalent to around 29% of the total
impressions delivered direct from the
Vote Leave Twitter account during the
referendum (2).
44. Overview & Key Questions
Period of Analysis: (1.1.2016 to 23.06.2016)
● Analysis of Russian Bots as revealed by US lawmakers (2,752 accounts) and 45 others, as
revealed by BuzzFeed
● This is an analysis of a small segment of discovered Russian bots (City University, London found
13,500 potential bot accounts)
● Who are the primary active bots seeding and sharing content about Brexit in the global
twittersphere?
● What is being shared about this topic both within the core community and beyond?
● What was the extent of influence of these bots on the results of the referendum?
45. Overview
● Total number of posts: 2,598 posts
(number of posts related to Brexit by
Russian bots in all languages)
● The word clusters (left) provide a
snapshot of the most commonly
discussed topics related to Brexit by
Russian bots between 1 Jan 2016 and
23 Jan 2016
● Combined, these tweets created
over 10 million total potential Twitter
impressions
46. (Day of Brexit) Various Russian bots tweeting.
Buzz chart: Jan 2016 - 23 June 2016
47. Mix of pro-Remain and
pro-Brexit hashtags
Top 10 most popular hashtags used by the Russian bots
48. Most mentioned users by the Russian bots 1/2
German national and international television
news service
German national weekly newspaper
Editor writing for ZEIT Online
President of the @EU_Commission
German social media user
49. Most mentioned users by the Russian bots 2/2
Donald Tusk, President of the European Council.
German finance reporter at Welt
(German national daily newspaper)
Bloomberg: business news (New York/ International)
German politician and Member of the European
Parliament from Germany
Ex-prime minister of UK
51. Most mentioned users by the Russian bots: analysis
● Within the top ten most mentioned users by Russian bots, 60% are German.
● Out of the remaining 40% of mentioned users, 10% of users are British. The one
British user is the UK prime minister at the time of the referendum, David Cameron.
● The other 30% of users are American and European.
● Rather than the bots generating new content about Brexit they are mostly
retweeting German users’ tweets about Brexit. As a result, several of the tweets
are in German.
53. Most retweeted Tweets (2)
● All accounts behind the top ten most retweets by Russian bots
have been suspended by Twitter, hence likely to be bots
themselves. Also, all accounts are from Germany that can be
geographically located.
● 70% of the top retweets are by the user '@blackjackweiner’. There
is a limited insight into this account given its suspension, however
their profile says that they joined Twitter in December 2015.
● Consequently, top retweets are equally German dominated as with
most mentioned users, and again, for inexplicable reasons.
54. Summary
The total number of posts related to Brexit by known Russian bots is
2,598 and the total number of tweets related to Brexit by all users
within this period is 22,110,853.
Consequently, from this narrow segment, Russian bots are responsible
for 0.0117% of tweets related to Brexit (in all languages).
The City University, London research found 5% of all EU referendum
tweeters were either deleted or recycled with a new name
56. Overview
● Total number of posts: 999 (number of
posts related to Brexit by Russian bots in
English)
● The most common terms are “brexit”
(40% of posts) and “eu” (27% of posts)
57. (Day of Brexit) Same as all languages
Key Drivers of Conversation
58. Top 10 most popular English hashtags used by the bots
59. (Same as all languages)
(Same as all languages)
(Same as all languages)
Latest comment, analysis and discussion from the British
newspaper, the Guardian
U.K. news, blogs and original content offering coverage
of various topics
Most mentioned users 1/2
60. (Same as all languages)
British financial journalist
An American television cable and satellite news and
satire channel
MP, Former Prime Minister of Estonia
Latest from the Finnish Government and ministries
Most mentioned users 2/2
61. Summary
● The top ten most mentioned users by Russian bots in English are a mix of
British (40%), European (40%) and American (20%)
● The total number of posts related to Brexit by Russian bots in English is 738 and
the total number of tweets related to Brexit by all users within this period (in
english) is 19,852,148.
● Out of the total number of tweets by Russian bots, 34% are retweets (almost the
same as all languages)
● Consequently, this segment of Russian bots are responsible for 0.0037% of
tweets related to Brexit (in english). City University’s figures are significantly
higher.
63. Terminology
● Paid social - social media content that the delivery is paid for
● Organic social - social media that is shared by users because they want to show the content to
their followers (regardless of platform).
● Estimated impressions - this is a calculation based on industry estimates of how many people see
a piece of content posted either on Facebook or Twitter. For instance, John Doe has 100 friends on
Facebook and posts a link with an image for a political article, our estimate that 10 of John’s friends
would see this in their timeline. Joe Doe has 100 Twitter followers, our estimate is between 5 and
10 of John’s Twitter followers would be served any of John’s tweets.
● Potential impressions - refers to estimates of the number of people who could have seen a Tweet.
5 users each with 1,000 Twitter followers tweet a link, this would be 5,000 potential impressions, as
potentially 5,000 people could see the link.
64. Notes on research
● (1) PBS and France24 shared content, data from Buzzsumo.
● (2) Russian bots delivered over 10,000,000+ potential Twitter impressions during the EU referendum. This is
equivalent to around 29% of the total potential impressions delivered direct from the Vote Leave Twitter account
from 22 Jan 2016 until 3 June 2016. This is based on the total number of tweets from Vote Leave in this period (790) x
average number of Vote Leave followers during this period (44,142) based on a follower count of 34,872 on 22 Jan
2016 and 49,983 on 3 June 2016 (we used a straight line estimate).
● (3) Based on an average mean Facebook user having 338 friends according to Pew Research. Impressions is the
maximum potential number of people who see a post, based on the number of Facebook shares times by the
average number of Facebook friends per Facebook user. The real figure is likely to be significantly less, but we
cannot tell reach as each post by each Facebook user could lead to a different number of impressions.
● (4) 39,310 is calculated using the higher figure for ‘likes’ and ‘follows’ for each Facebook page (there can be overlap
between likes and follows, but also no overlap, so using the larger figure of the two gives a conservative estimate of
actual people having engagement with an account - eg. one user may like and follow one page, another user may
just follow that page). The total across all the pages is divided by estimated organic reach which at the time of the
EU referendum, a conservative estimate is around 3%. Many Facebook pages with an engaged audience saw
considerably higher reach, often over 10%, but we are using a lower, more conservative figure for Facebook pages.
65. Notes on research
● (5) We are using a reach figure of 33.4 for each individual organic post. This figure is based on the average
Facebook user having an average number of friends of 338 and an estimated organic reach of 10% from link
posts by individuals giving 33.4 as a reasonable estimate of the number of people who would see each post
shared by an individual. Many of the posts will have been shared by groups or pages with large numbers of
followers giving bigger reach than 33.4; this is offset by Facebook posts reaching people outside the UK though
the Leave campaign organic figures include non-UK nationals (it is difficult to separate this data). Facebook
page sizes based on figures from late June 2016; eg. Vote Leave (554k).
● (6) The figure for the value of Russian media coverage of the Leave campaign is based on 261 Brexit articles
published by RT between 1 January 2016 and 24 June 2016, of which 98 had a pro-Leave slant and 163 articles
published by Sputnik International which had a clear pro-Leave slant according to our researchers. We will
publish the full list for transparency. Each RT article has a media value of £5,681 and each Sputnik article £4,886,
according to media monitoring organisation Kantar.
● (7) We use an average of $1.45 USD to GBP for the period 1 January - 22 June based on historical data. We
exclude the major fall in the currency after the result of the referendum was known.
66. Notes on research
● (8) The figure for the value of Twitter impressions is based on the Twitter API search result of 200,000,000
impressions for content on RT/Sputnik converted into delivered impressions using a 5% potential to delivered
impression conversion and a CPM of $7 (the 2016 Twitter global figure). The global figure of $7 CPM is arguably
lower than the UK CPM in 2016, especially during the EU referendum. The conversion of 5% is a medium figure
based on a low number of accounts with over 100k followers sharing the content, which tallies with our analysis
of the shares. We give two figures, one at 5% conversion, one at 10% conversion.
● (9) Ruptly had 84,781 likes as of 6 October 2016 and 28,099 a year before on 31 October 2015. Based on straight
line follower growth we estimate Ruptly had 60,000 followers in June 2016. With 72 videos posted on their
channel plus the organic shares we estimate their total Facebook impressions to be 10,850,000.
● (10) The top 100 articles were analysed for their coverage of either the Leave or Remain campaigns; our
researchers considered how a reasonable person would read the coverage and draw a conclusion as to
whether the coverage was positive (or negative) about either the Leave or Remain campaigns, or was neutral
about the campaigns. Articles negative about the EU are included in the Leave coverage. An article which
highlighted an announcement by a particular campaign - without coverage for the other campaign - is
categorised as coverage for that campaign.
67. Notes on research
● (11) 'Dark Posts' are social media posts which are promoted through the social media platform’s advertising
tools to target audiences. The posts are referred to as ‘dark’ as they cannot be seen by users unless they have
been specifically targeted. These posts cannot be seen on the advertiser’s social media account at all.
68. RT most influential
sharers of content by
Klout score
RT Leave articles
16,545 Total Posts
120 million total potential
impressions
Author Name
@TRobinsonNewEra Tommy Robinson GB
@georgegalloway George Galloway
@Trump_Videos Deplorable Covfefe
@End_of_Europe Stephanie Cisnero
@janimine Jani
@V_of_Europe Voice of Europe
@CherylShuman Cheryl Shuman
@COLRICHARDKEMP Richard Kemp
@RayJoha2 Raymond Johansen
@_HankRearden #Bossy Hank III
@davidicke David Icke
69. Sputnik most influential
sharers of content by
Klout score
3,389Total Posts
14 million total
potential
impressions
Author Name
@SputnikInt Sputnik
@PrisonPlanet Paul Joseph Watson
@TRobinsonNewEra Tommy Robinson GB
@End_of_Europe Stephanie Cisnero
@V_of_Europe Voice of Europe
@YourAnonGlobal Anonymous
@MikkiL Mikkil #BRINO #Brexit GBUS
@WellyTopping Welly Topping
@HouseCracka General Deplorable
@DonaldTrumpWall DonaldTrumpWall