It begins with a canal project in Afghanistan and a cafe worker called Sophia, but ends by explaining how 91% of likes on UAE Crown Prince @MohamedBinZayed's tweets are created by at least 11,000 bots/fake accounts #disinformation#UAE
2/ So what alerted me to this weirdness. Firstly, a series of strange tweets about a UAE project in Afghanistan to build a canal.
'Kate Johnson' and 'Jessica Anderson' were really amped about the project.
3/ What was odd about Johnson and her ilk is how specific they were. Sophia Moore, for example, claims to work in tea junction cafe in Ajman...It's a real place, but who wants to bet Sophia doesn't work there?
...even if she does have to work to 'have money for fashion'
4/ Another example is Lara Adams, who works in a Garden Centre in Dubai, but for some reason uses a stock photo that is popular with dentists. There were dozens of accounts like this, united by their love of the UAE, canal projects in Afghanistan, and stock profile photos.
5/ Anyway, these aren't the particularly droids I'm looking for. However, I noticed that one of them, *ahem*, Megan Jones, had a weirdly high number of likes - 901 - on her tweet. In the industry we call that 'loads'. Also no retweets or replies. Odd right?
6/I clicked on the likes, and immediately noticed that many of the accounts gave creepy AI image vibes and/or had the same identical cropping.
Now these are the droids I'm looking for...
7/ Things started getting really weird when I found two accounts with identical bios. What are the odds of two people who are phd holders in the management of UAE's nuclear power plant who are also 'Swiss addicts'.
Also, wth is a Swiss addict?
8/ So I explored what these accounts were liking. All of them liked tweets by @MohamedBinZayed . Many of them boosted tweets critical of Al Jazeera, supportive of normalization with Israel, and against Qatar and Turkey. The majority though were promoting the UAE and MBZ
9/ So I wanted to an idea of just how many of these fake bois there were. I downloaded the likes and RTs of an MBZ tweet about Covid, which had around 6600 likes and 9000 RTs. I then analysed an 8000 account sample using the creation date method... #disinformation
10/ It revealed something striking. 50% of the accounts were created in just 25 days, bear in mind Twitter has been around since 2006. As the graph below shows, 41% of the accounts who liked/RT'd MBZ's tweet were created in just one month, June 2020!
11/ Based on the anomalies here, it's safe to hypothesize that the majority of accounts created after May 2020 are also sockpuppets. There were other patterns too, namely they all had similar name structures, follower/following ratios and total tweets. #disinformation
12/ Based on this, a conservative estimate is that 91% of the likes of that MBZ tweet are generated by bots.
I looked for another tweet that would have fewer genuine likes than MBZ. Who did I find the bot network boosting, but NSO/Pegasus and dictator apologist Irina Tsukerman!
13/ 8350 likes and 7 retweets is just bizarre. Anyway, I analyzed the likes, and in Irina's case the percentage of fake likes is a mind-blasting 99.7%, that's around 8333 bots! I combined these accounts with the known bots from MBZ's tweet and it comes to around 11000 unique bots
14/ Some points > It is not clear which of the accounts being boosted actually knows if they're being boosted. Pretty sure MBZ doesn't run his own account.
I believe the network operates to strategically boost some posts in certain multipliers. Not every account in the
15/ network will retweet identical posts. So even if an MBZ post has say, 4k likes and not 6k, a large number of them will be fake. This is probably to give the illusion of some level of authenticity...
16/ To sum up. There are at least 11,000 fake accounts that post pro UAE/MBZ content or content reflecting UAE foreign policy, including support for Israel, and flak against al Jazeera, Turkey and Qatar. It's not clear who runs them, but their stance is consistent! It's also
17/ super creepy that bots are boosting other bots. Botception! Anyway, happy to answer any questions you have. Hope you enjoyed today's instalment of bot wresting! #disinformation
• • •
Missing some Tweet in this thread? You can try to
force a refresh
🧵1/ I analysed the headline and lead paragraph of 536 English news articles including the terms "Maccabi" + "Amsterdam" and classified them using Claude 3.5 Sonnet to determine how many framed Israelis as victims or non-Israelis as primary victims (as well as both).
2/ The results are fairly striking. 65% of articles frame Israelis as the victim, while only 5% frame Non-Israelis as victims. 24% are neutral while 9% framed both groups as victims. Quite clear the media emphasised violence as anti-Israeli and antisemitic, especially early on
3/ There isn't much evidence too of corrective framing at this point, although a small increase in neutral framing a week after the incident. Israeli victimhood was categorised as emphasis of violence initiated by non-Israelis, and focus on anti-Israeli or antisemitic violence
🧵 1/ Part of understanding what is going on in Amsterdam is also to understand the coordinated anti-Arab, anti-Muslim and anti-immigrant campaigns run with huge amounts of money targeting Europe. Here's a short private Eye article about an investigation I did with @SohanDsouza
2/ Here's a write-up by @karamballes on the campaign in @BylineTimes "Disinformation Campaign on Social Media Reached More Than 40 Million People – but Meta ‘Alarmingly’ Hasn't Revealed the Culprits' bylinetimes.com/2024/08/30/qat…
@karamballes @BylineTimes 3/ ...How a covert influence campaign helped Europe’s far right
Our findings about the shadowy multi-platform operation attacking Qatar and stoking Islamophobia to further its far-right agenda in Europe and beyond call for immediate action. aljazeera.com/opinions/2024/…
🧵🚨1/ This is nuts. After mysteriously deleting a package covering the Amsterdam protests, Sky News have put up a new version. The new version completely changes the thrust to emphasise that the violence was antisemitic. See the opening screenshot change below
2/Even the tweet accompanying the video has changed. It has explicitly shifted from mentioning anti-Arab slogans to removing the phrase "anti-Arab" and using antisemitism. It also removes mention of vandalism by Israeli fans. An extremely clear editorial shift!
3/ They have also inserted into the video, right after the opening footage of Dutch Prime Minister condemning antisemitsm. This was not in the original video.
1/ If you break down the BBC's live reporting of what happened in Amsterdam, you can see the disproportionate attention it pays to Maccabi fans and Israelis as victims, with far less attention paid to the actions of Maccabi fans. Here are the sources interviewed.
2/ In terms of mentions of Arab, Dutch or other Ajax fans, there is very little emphasis on Arab safety, with the majority of coverage focused on Maccabi fans as victims. There are vox pops with fans, but very little interaction with non-Maccabi people.
3/ The language used to describe the attacks on the Maccabi fans is also much stronger, ranging from pogroms to brutal and shocking. Similar terms aren't use for the anti-Arab racism.
🚨1/ This New York Times piece is wild. Let's go through it.
Firstly, the lede is an emphasis that attacks in Amsterdam were based on antisemitism, yet it cites no evidence of this, but DOES cite evidence of anti-Arab chants.
2/ The claims of antisemitism are based primarily on the Prime Minister of the Netherlands, who tweeted that the attacks were antisemitic. Note - the Dutch Prime Minister didn't call out anti-Arab or anti-Palestinian racism from Maccabi fans.
3/ The piece links to an Amsterdam police statement to talk about the violence - although the police statement doesn't mention anything about antisemitism.
🧵 'At least 1,800 bots on the social media site X are promoting the controversial choice of Azerbaijan, a major oil and gas producer, to host next month’s ...#COP29, according to a new analysis shared exclusively with The Washington Post".
2/ The analysis by Marc Owen Jones, an expert on disinformation at @NUQatar, focused on roughly 2,800 X accounts that collectively sent around 10,800 tweets, retweets and replies about the conference between Oct. 17 and Oct. 24.
3/ Detection
73% of all accounts active in sample created in the space of 3 quarters in 2024.
Conservative estimates suggest 66% (1876) accounts in the sample are fake (bots) based on activity over the past week