Newly identified bot-like accounts are generating billions of impressions, sowing division and spreading disinformation on X

31 July 2024, London – 45 accounts that appear to be bots on X have generated over 4 billion views while amplifying racist and sexualised abuse, conspiracy theories and climate disinformation, according to a Global Witness investigation released today.

Global Witness identified the bot-like accounts that collectively produced around 610,000 posts in the period from May 22 to July 22. Despite sampling across users posting hashtags supporting the four largest political parties (by vote share), the majority of bot-like accounts the investigators uncovered expressed support for Reform UK.

Identified from their activity ahead of the UK general election, a number of the bot-like accounts subsequently pivoted to emerging global events, including anti-migrant protests in Ireland, the assassination attempt against Donald Trump, and Joe Biden stepping down from the US presidential election race. They often responded to these events with racism, gendered disinformation and conspiracies.

Ava Lee, Campaign Leader of Digital Threats at Global Witness, said:

“In a moment when everyone is worried about democracy, it is shocking how easy it has been to find accounts that appear to be bots spreading division around the UK vote, and then to watch them jump straight into political discussions in the US and Ireland, frequently responding with hate and conspiracy.

“Access to timely, accurate and reliable information is crucial for all democracies, and it’s needed now more than ever in the run up to the US presidential election.

“In light of our findings, we are calling on X to increase its moderation efforts and get better at enforcing its own policies against inauthentic activity. We’re relying on them to protect our democracies from interference.”

Although there is no evidence that UK political parties are paying for, using or promoting these bot-like accounts, they include 28 accounts posting in support of Reform UK, 14 accounts in support of the UK Labour Party, and a small number of accounts in support of the Liberal Democrats and Conservatives.

Tracking them after the UK general election showed their activity was generally split along partisan lines, with 14 of the bot-like accounts that posted in support of Reform UK sharing the hashtag #ClimateScam in the post-election period.

The investigation also shows that a number of the bot-like accounts that posted prolifically during the UK general election move to rapidly respond to emerging new topics amplifying divisive content. These include:

Trump assassination attempt: The investigation documents the sharing of conspiracy theories on the assassination attempt on US presidential candidate Donald Trump, with some accounts previously supporting Reform UK suggesting that Joe Biden was responsible and some accounts supporting Labour questioning whether the assassination was staged by Donald Trump.
US presidential race: After US president Joe Biden announced he was stepping down from the presidential race, a number of bot-like accounts increased their discussions of Kamala Harris, including sharing negative racist and sexualised comments and known gendered disinformation tropes.
Anti-migration: The investigation documents bot-like accounts participating in online discussions relating to the recent anti-migrant protests in Coolock, Dublin, including conspiracy theories and use of hashtags such as #IrelandbelongstotheIrish and other xenophobic sentiments.
Climate change denial: Sharing of climate conspiracies, including using hashtags such as #ClimateScam or #geoengineering.

Global Witness wrote to X to give them the opportunity to comment on these findings, but they did not respond to our findings.

Notes to editor:

Methodology

Sampling:

We focused on accounts promoting the four political parties that captured the largest vote share in the UK election: Conservative, Labour, Liberal Democrat, and Reform UK
We took a sample of posts using #voteconservative, #votelabour, #votelibdem, #votereformuk from the six-week period between the election being called and taking place (May 22-July 4) and examined whether accounts using those hashtags show bot-like behaviour
It’s often not possible to know with certainty that an account is part of a coordinated bot network or troll farm

Identifying bot-like behaviour:

Our approach focused on surfacing accounts which showed evidence of inauthenticity, automation or disproportionate participation in political conversations online
We examined indicators that we are calling "red flags", such as accounts that post enormous numbers of posts per day or share very little original content
Accounts with a series of "red flags" were reviewed manually to examine if there was evidence of their identity, personal life or interactions which were very likely to be authentically human

Red flags:

The account is a prolific tweeter and posts in high volume, likely to be aiming to drive online conversations:

The account has tweeted more than 60 times a day on average over the studied period
And if so, the account has tweeted more than 144 times a day on average over the studied period
The account posts in multiple languages (has posted more than once in more than 3 languages)

The account posts a low amount of original or high-quality content and predominantly reposts (reposts are easier to automate, and this may generate some following but is unlikely to lead to mass followings):

The account reposts other accounts’ posts more than 90% of the time
The account posts original posts (not reposts, replies or quote posts) less than 10% of the time
The account has posted duplicate content of more than 1 word

The profile was able to be set up quickly in order to respond to an event:

The account’s handle ends in a long string (greater than 5) of apparently random numbers, suggestive of a handle being used that was generated automatically by X
The account was created in the last 50 days

The profile indicates possible deception:

The account name and handle do not match in a way that is suspicious

The profile has indications of potential coordination with others:

An account’s posts receive a suspiciously consistent number of likes for its posts
An account’s posts receive a suspiciously consistent number of reposts for its posts

Political affiliation:

We determined the political affiliation of each of the accounts by reviewing how often they had used a partisan hashtag, along with reviewing the accounts for explicit or implicit support for a political party in its account bio, cover and profile photos, posts or media content

Endnotes

We used Information Tracer to help with our analysis
We do not have any evidence to suggest that any UK political party is paying for, using or promoting bots as part of their election campaigns
Figures are estimated, and may vary where accounts/posts have been deleted

Newly identified bot-like accounts are generating billions of impressions, sowing division and spreading disinformation on X

Contacts

Notes to editor: