We investigated an online influence network in the UK that wants to see Boris return as prime minister
On Monday, September 5, we will know who Conservative Party members have chosen to replace Boris Johnson as the UK’s prime minister. Public attention has been focused on the campaigns of the two candidates, Liz Truss and Rishi Sunak. But we have found a third campaign, an insurgent effort using questionable online tactics and Trump-style vitriol, that may have a more profound impact on UK politics in the long run.
Monitoring online activity around the Conservative Party’s leadership election, we found what looked suspiciously like a coordinated network on Twitter promoting a campaign to keep Johnson in power. We were drawn to the network due to its use of Trump-style language; referring to former chancellor Rishi Sunak as a “traitor” and a “snake”, using hashtags such as #NeverRishi, while at the same time, praising Truss, who has been careful to maintain the appearance of loyalty to Johnson. The tone of the messaging promoted by the network also had a distinct Trumpist flavour in the way it vilified mainstream media; one of the network’s most popular hashtags was “#scummedia”, commonly used by the former president’s fans in the US when he is firing broadsides against the likes of CNN. Given all this, it is perhaps not that surprising that the United States was the second most common country of origin for accounts in the network (after the UK), and some of those American accounts display their support for Trump in their bios.
An initial manual search through the tweets associated with the network suggested its main aim is to force the Conservative Party to allow Johnson to stand in the leadership election, effectively allowing the membership to overturn his removal by Conservative MPs. One domain in particular, Conservative Post (CP), seemed to feature prominently in the network’s tweets. A quick look at the website makes it clear CP is a news outlet campaigning for Johnson’s reinstatement as prime minister; the premise of which is based on a rule in the Conservative Party’s constitution that allows grassroots demands that attract 10,000 votes in a petition to be put to a membership-wide vote. The website is heavily promoting a petition to get the signatures necessary to force Johnson on to the ballot. Vocal backers of the petition include Peter Cruddas, a billionaire businessman appointed co-treasurer of the Conservative Party in 2011, who Johnson made into a life peer of the UK’s upper chamber of parliament, in 2020. News reports suggest that senior Conservative Campaign HQ (CCHQ) officials seemed a little exasperated with the campaign, dismissing it in its early stages.
We have suspected for some time now that online manipulation has been used to promote Johnson’s political career. In 2019, during the general election that gave the Conservative Party a landslide victory, online manipulation came to Johnson’s rescue when a video surfaced in the final days of the campaign showing a sick boy having to sleep on the floor of Leeds Hospital due to under-resourcing of the UK’s health service. Although media coverage at the time (see here and here for examples) identified that disinformation techniques had been employed, there was no further follow-up on how it was done, or who may have been responsible.
Based on these observations, we decided the network warranted further investigation. We used the network’s most popular hashtag, #BackBoris, to dig into its methods. We found plenty of indications of coordination, but not much that was as cut and dry as what you might see in Africa or the Middle East.
To obtain our dataset, we captured all the tweets using the hashtag between July 7 and August 7. The total came to 62,395 tweets from 9,621 accounts. Social media listening software suggests a potential reach of 14.7 million; this figure should be treated with caution as it does not represent an accurate calculation of all the people who saw the hashtag. It does however suggest that over the month, the #BackBoris campaign would have featured heavily in the timelines of the 180,000 to 200,000 people eligible to vote. Unsurprisingly, Conservative Post was the most commonly shared source of information; much more so than larger and more established right-wing news outlets such as the Telegraph or television network GB News.
To analyse the way the accounts were using the hashtag, we applied the Nimmo Coefficient of Traffic Manipulation (CTM), developed by Meta’s Ben Nimmo to this dataset. CTM compares traffic between different hashtags, assigning a rating based on the average number of tweets per user, percentage of retweets, and the percentage of total traffic that is made up by the 50 most active users. A high score indicates a hashtag is being spread through the use of manipulation techniques. The #BackBoris hashtag was rated at 29.75. Control samples (generic hashtags such as #WorkOut or #Thursday) tend to achieve a rating of 12. The rating of the #BackBoris network shows a high degree of manipulation.
Contributing to the high score was the fact that over 80% of the tweets in the dataset were retweets, which is unusually high and points to the use of bots (or intensely motivated individuals) working to make specific hashtags trend.
When we examined the 10 most active accounts in the overall dataset, we found that bot-detection software Botometer rated 7 of them as likely automated and/or inauthentic. The high rating reflected activities such as posting more than 150 times a day, almost exclusively retweeting (rather than posting original content) and posting consistently throughout the day (ie little to no breaks for work, sleep etc.)
When we looked at the top 10 accounts promoting links from the Conservative Post website, we found similar bot-like behaviour. Common themes in the bios of these accounts included, along with pro-Trump hashtags, indications of anti-vaxx (#vaccineinjured), climate denialism (#cancelnetzero), and anti-BBC (#DefundTheBBC) sentiment.
We also found authentic-looking accounts attempting to coordinate activity in a highly organised, consistent and strategic manner.
The network graph shows four clusters within the #BackBoris network. At the centre of the clusters is a node acting as a coordinator (or, in one case, a small group of accounts acting as a node). The nodes appear to be run by real people coordinating the online activity of others in the network.
The screenshots below show how one of these nodes (also the most retweeted account in the wider network) coordinates users in order to amplify pro-Boris hashtags. This involves directing other users’ activity, drawing attention to pro-Boris accounts or compiling Twitter lists of accounts for others to follow. Other users focus their content on particular strands of strategic pro-Boris messaging (anti-Sunak, pro-Truss, anti-Labour).
Overall a relatively limited number of inauthentic, bot-like, accounts were used to retweet en masse; while real-life users conducted the “higher-quality” task of engaging with users and coordinating their activity. We didn’t find extensive use of automated software. But we did find that a group of highly motivated, organised and consistent individuals spent a lot of time engaged in activities designed to promote Johnson’s cause.
Although Conservative Post and its campaign to re-elect Johnson as prime minister are clearly popular with the network, we could not prove a substantial link (e.g. funding or technical infrastructure) between the website and the network promoting it so enthusiastically. However, web analytics tool Similarweb shows that over the last two months, 71% of the Conservative Post website came from social media, and 61% from Twitter alone, suggesting that the #BackBoris network is a key driver of its online relevance.
A combination of authentic and inauthentic activity pushes the network’s messaging. “Inauthentic” generally means the use of automation software and other technical means. This is categorically banned by platforms. “Authentic activity” is taken to mean that humans are acting together to push their agenda on social media: Here, the rules are less cut and dry. Twitter, for example, calls such activity “social coordination” and says it is unlikely to act against activism and efforts to “speak truth to power… unless we can establish evidence of harm”. This means that the network we found, if investigated, might have some automated accounts closed, but would largely be allowed to continue its efforts to promote Johnson’s career.
To be clear, we could have investigated the network further, by, for example, assessing the CTM rating for other hashtags used by the network, assessing the bot rating of the various accounts
As always with influence operations, the key question is “so what?; what has all this effort actually achieved?”
Johnson resigned on July 8. A YouGov opinion poll from a few days before showed that 69% of British voters wanted him to step down. Amongst those who voted for the Conservative Party in 2019, the number was 54%. Less than a week after Johnson’s resignation, Conservative Post set the tone of its campaign with a Twitter poll (tweet included below) that claimed 503 out of 508 Conservative Party members (ie those who ultimately decide the winner) would vote for Johnson if he was on the ballot. The Tweet was continuously retweeted for a week, which is in itself unusual.
Over the next few weeks, the network set to work advocating for members to petition the Conservative Party to include Johnson on the leadership ballot, frequently drawing on articles from Conservative Post to demonstrate the legal, moral and democratic legitimacy of its campaign. On August 13, an Opinium survey reported in the press found that if offered the choice, 63% of Conservative Party members would rather Johnson back in charge instead of seeing Sunak or Truss become prime minister.
The media analysis suggested that support for the “Johnson option” was due to a lack of enthusiasm for the remaining two candidates (who had been whittled down from a starting line up of 8 in the early days of the campaign) rather than an underlying, grassroots desire to have Johnson back as prime minister.
Conservative Post, however, painted the survey’s results as proof “ the level of support he [Johnson] enjoys within the Conservative Party membership has not waivered”.
By August 25, the Telegraph reported that Conservative Campaign Headquarters (CCHQ) had received 8,700 signatures from members calling for Johnson to be included on the ballot.
It would be a stretch to put Johnson’s apparent popularity bounce down to the efforts of Conservative Post and the online activity of the #BackBoris campaign, but they did exploit the change in sentiment to mobilise thousands of Conservative members in the hope that 10,000 signatures could trigger Johnson’s return to power. Also, we cannot dismiss the possibility that the scale of the consistent messaging put out by the campaign affected the way members’ saw the race. A significant proportion of those asked mentioned “loyalty” as a reason for supporting Truss while distrusting Sunak.
This is a surprising achievement for a marginal political news outlet with a Twitter account of just over 7,000 followers, and Facebook following of 11,000. How did it manage it? Based on the finding that most of Conservative Posts’ traffic comes from social media, it is worth looking at the network’s content.
The mass retweeting and frantic coordination between committed online activists drives a very specific type of messaging. While the website’s tone is upbeat and positive, the network uses the sort of divisive messaging that psychologists Rathje, Van Bavel and van der Linden have shown is highly effective at attracting attention on social media. (see the screenshot below)
Beyond using “us-vs-them” messaging, the network also uses Trump-like conspiracy; specifically the idea that hidden powers are trying to subvert the “will of the people”. In the case of the #BackBoris network, the nefarious actor is often the EU and other Conservatives (particularly Sunak) are its minions.
Echoing Trump’s attacks on Vice President Mike Pence and his depiction of insufficiently supportive Republicans as “RINOs” (Republican In Name Only), members of the network frequently blame “backstabbing” Conservative MPs for staging an “anti-democratic coop d’etat” (sic) against Johnson. The party’s 1922 Committee, which oversees election rules, also finds itself in the network’s sights for its perceived reluctance to have Johnson stand again. Posited against these shadowy actors is the pure, unsullied will of the people – as expressed by Conservative Post and its online supporters. Sound familiar? (see example of a conspiracy-minded tweet below)
Of course, political influence operations are nothing new. Traditionally they are waged through recognised mediums, such as mainstream media outlets, ads, endorsements etc. However, a situation such as the leadership race of a ruling political party – where a relatively small number of fairly like-minded people get to decide who rules the entire country – presents an ideal scenario for online persuasion. Skilled and experienced operators know how to make sure a relatively small but homogenous group online keep seeing their messaging day after day. Whereas mechanisms exist to monitor campaigning activity using traditional tools, governments, even well-resourced ones such as the UK, have limited capacity to investigate potentially hundreds of social media networks that pop up around an election and understand who is paying for them.
Successful online political operations tend to be quickly adopted and adapted. It seems in the UK, someone thinks it is worth copying Trump’s playbook to promote the prospects of Boris Johnson in British politics. The risk is, if the effort is seen as effective, we can expect to see many more such anonymous networks springing up to use toxic messaging for short term gain; and there’s little in the way of already existing rules to stop them.