Calls for violence and antisemitic abuse targeted at British Jews on X platform following Manchester Synagogue Attack

Prominent British Jewish organisations and individuals who posted expressions of grief, horror and frustration following the attack at the Heaton Park Hebrew Congregation Synagogue on October 2^nd were targeted with antisemitic abuse on the X platform.  

CCDH researchers analysed the replies to X accounts belonging to prominent British Jewish individuals and organisations who posted about the attack and found severe instances of antisemitism or calls for further violence.  

The list of X accounts analysed includes prominent Jewish individuals such as religious leaders, television personalities and journalists alongside Jewish organisations that work to support the Jewish community.  

In response to posts from the accounts following the Heaton Park Synagogue attack, researchers identified a range of antisemitic abuse including holocaust denial, calls for Jews to be deported, dehumanising language, celebration of terrorism and calls for further violence.  

Examples of hateful and violent posts targeting the accounts include the following: 

• “Shame they didn’t get more”ⁱ 

• “It’s time for the humankind to eradicate the Sinagogue of Satan.”ⁱⁱ 

• “Every time a jew dies, the world becomes a better place.”ⁱⁱⁱ 

• “Shut up jew. Always sticking your beaks in everyone’s religion.”^iv 

• “Wish they burned them all alive in the darkest nastiest people on Earth!! Criminal committing crimes since Creation of the wolrd!! #jewshit #SupportPalestine #DemolishIsrael”^v 

• “good as they should. don’t stop till israel and all jews are gone off this blessed earth.”^vi 

• “Only two jews died. That’s the saddest thing about your little synagogue party. You got off easy.”^vii 

These hateful replies were to posts made by 13 of the most prominent British Jewish individuals and organisations. Researchers identified them by analysing replies to posts containing relevant keywords in the five days after the attack, such as “Manchester”, “synagogue” and “Yom Kippur”.  

Researchers amassed a dataset of 15,959 posts and analysed 171 hateful replies, identifying at least five instances for each of the 13 accounts studied. The hateful reply posts were all live on the platform as of the time that the analysis took place. 

X is failing to uphold its own policies on hateful conduct and violent speech  

The X platform has failed to effectively moderate any of the replies CCDH identified, despite violent and hateful content being explicitly prohibited under X’s rules.  

Specifically, between the platform’s “violent content” and “hateful conduct” policies, the platform states that it does not allow the following content: 

• Inciting, promoting or encouraging others to commit acts of violence or harm.^viii (Violent content policies) 

• Wishing, hoping or expressing desire for harm, including hoping for others to die, suffer illnesses, tragic incidents, or experience other physically harmful consequences.^ix (Violent content policies) 

• Threats to inflict physical harm on others, including threatening to kill, torture, sexually assault, or otherwise hurt someone.^x (Violent content policies) 

• Targeting individuals or groups with content that references forms of violence or violent events where a protected category was the primary target or victims, where the intent is to harass,” including the holocaust.^xi (Hateful conduct policies) 

• Directly attacking other people on the basis of religious affiliation and inciting fear or spreading fearful stereotypes about a protected category.^xii (Hateful conduct policies) 

Posts identified were in contravention of X’s rules, indicating that these policies are not being enforced effectively and raising serious concerns about the risk of further violence. 

At the time of his acquisition of X, Elon Musk promised advertisers that the platform “obviously cannot become a free-for-all hellscape” and “must be warm and welcoming to all.”^xiii He also promised to make X into “a common digital town square, where a wide range of beliefs can be debated in a healthy manner, without resorting to violence.”^xiv 

The UK Online Safety Act requires that social media platforms like X enforce their own platform rules and have systems and processes in place to identify and remove illegal content.^xv Examples contained in this analysis raise serious doubts about the X platform’s adherence to those legal requirements. 

Examples of posts promoting violence 

This post is in response to a TV personality expressing prayers for the victims of the attack.^xvi 

This reply is to a post by a TV personality praying for the victims of the attack.^xvii 

This post is in response to a former newspaper editor voicing anger over antisemitism.^xviii 

This reply post is in response to a post by a newspaper columnist.^xix 

This post is in response to a post from a Jewish organisation in which they voice frustration at Tommy Robinson’s official invitation to Israel in the wake of the attack.^xx 

This post is in response to a post from an author expressing frustration about pro-Hamas narratives in the context of the Manchester attack.^xxi 

Examples of antisemitic posts 

In a reply to a columnist and broadcaster, this post characterizes the holocaust as a “holiday camp”.^xxii 

This is a reply to a post by a Jewish organisation in which they voice frustration at Tommy Robinson’s official invitation to Israel in the wake of the attack.^xxiii 

In a reply to a post by a journalist, this post claims “jews are an infestation” and calls them “parasites”.^xxiv 

This is a reply to a post in which a campaign group express their grief in reaction to the attack.^xxv 

This post is in response to a post by a TV personality in which he said he is praying for the victims of the attack. 

This is a reply post to financial journalist in which he expresses his grief of the attack and his pride in being British. ^xxvi 

None of 100 top-viewed ‘false flag’ conspiracy posts displayed Community Notes 

Out of 100 posts on X spreading the conspiracy that either Israel or Jews were behind the Manchester attack, none displayed Community Notes. Together, the posts accrued 568,708 views. 

This conspiracy took various forms, with posts often describing the attack as a “false flag”. Some claimed that the motivation behind the attack was to gain sympathy for the war in Gaza or to act as a distraction. Others pointed towards the name of the suspect as supposed evidence of the conspiracy. 

Examples of posts promoting the conspiracy include: 

• “The Manchester Synagogue attack is a false flag operation orchestrated by Israel mossad to gain sympathy.” ^xxvii 

• “His name was Jihad The Syrian?? Another Israeli false flag right after Britain recognised a Palestinian state.”^xxviii 

• “#SynagogueAttack Manchester “terrorist” hoax instigated by CST security and our government. Footage being cut by media so you cannot see the CST actors in place – here is the full footage. Israel Zionists have infiltrated the UK Government and Police Force. Wakey wakey.”^xxix 

• “Another Jewish false flag in Manchester.”^xxx 

• “At a time when when [sic] #GlobalSumudFlotilla nears Gaza & IOF terrorists start seizing the boats & abducting the activists violently, attracting outrage & protests in parts of the world, suddenly an ATTACK on synagogue in Manchester. Perfect script by #Mossad.”^xxxi 

By promoting this unfounded and sinister conspiracy, the posts echo historic antisemitic narratives that scapegoat Jews for tragic events and that blame Jews for seeking to manipulate world events to further their own plans.^xxxii 

Researchers checked each of the 100 posts promoting the conspiracy, finding that none of them displayed Community Notes. X’s Community Notes is a crowdsourced fact-checking system that allows users to provide context, clarification, or corrections on posts.^xxxiii The notes are only shown to users if they receive “enough” votes from contributors with “different points of view”.^xxxiv 

This finding suggests that the Community Notes system is failing to appear on antisemitic conspiracies, despite Musk’s claim that: “We’ve put maximum resources and attention behind Community Notes. So, if somebody tries to push a falsehood like Holocaust denial or something like that, they can immediately be corrected”.^xxxv 

Methodology: How we identified abusive replies 

Researchers used an AI-assisted process to identify a total of 171 abusive replies posted in response to prominent Jewish individuals and organizations on X in the wake of the terror attack which targeted worshippers at a synagogue on Thursday 2nd October. 

To do this, researchers created a list of prominent British Jewish individuals and organizations that had posted following the attack. The list was compiled by selecting prominent Jewish individuals and organisations from a wider list of accounts that had received large amounts of engagement on their posts relating to the attack. 

Using a licensed third-party tool, researchers identified 125 posts about the attack from the 13 accounts, including all posts that hit the below keyword query during the timeframe from the 2^nd of October to the 5^th of October. They then extracted 15,959 replies that were posted in response to those posts up to the 6^th October.  

(((Manchester) AND (synagogue OR crumpsall OR attack OR stab OR stabbing OR terroris* OR jews OR jewish OR jew Or “Yom Kippur”)) OR ((crumpsall) AND (synagogue OR attack* OR stab* OR terroris* OR jews OR jewish OR jew Or Israel OR “Yom Kippur”)))

These reply posts were filtered by utilizing OpenAI’s GPT-4o large language model to identify those that were likely to contain abusive antisemitic hate or violent content. Each post was assigned a score from 0 to 10 indicating the likelihood that it met the criteria.  

Comments assigned a high likelihood of containing antisemitic hate or violent content were then reviewed by researchers, who assessed the posts as to whether they were indeed hateful in this way. Posts were only included in the final dataset where they had been reviewed by two researchers, with both agreeing that the post contained antisemitic hate or violent content that was likely to breach X’s policies on “hateful conduct” or “violent content”. 

Data collection of reply posts took place on October 8^th 2025. Data collection for posts about the attack for the 13 accounts took place in two batches: once on October 3^rd 2025 and once on October 6^th 2025. Analysis of the reply posts took place on Thursday 9^th October, and all the 171 hateful reply posts were live on X on this date. 

Methodology: How we identified the 100 top-viewed false flag posts 

Researchers used an AI-assisted process to identify the 100 top-viewed posts that promoted the conspiracy that the terror attack which targeted worshippers at a synagogue on Thursday 2^nd October was orchestrated by Israel or Jewish people.  

To do this, they used a licensed third-party tool to identify 127,740 posts about the attack that hit the below keyword query during the timeframe from 2 October through 5 October.  

(((Manchester) AND (synagogue OR crumpsall OR attack OR stab OR stabbing OR terroris* OR jews OR jewish OR jew Or “Yom Kippur”)) OR ((crumpsall) AND (synagogue OR attack* OR stab* OR terroris* OR jews OR jewish OR jew Or Israel OR “Yom Kippur”)))

These posts were first filtered by utilizing OpenAI’s GPT-4o-mini large language model to identify and remove posts that were commenting on the event neutrally. This filtered the dataset down to 97,027 posts that were analysed using OpenAI’s GPT-4o large language model to identify antisemitic content, including conspiracy theories. Each post was assigned a score from 0 to 10 indicating the likelihood that it met the antisemitic criteria.  

To evaluate and calibrate the model, researchers sampled a stratified set of 200 posts for human labelling, with 100 examples that were deemed likely to contain antisemitism and 100 posts that were likely to not contain antisemitic content. Based on these labels, a decision threshold of 7 was chosen. At this threshold, the model achieved a re-weighted recall score of 46%, meaning it correctly flagged nearly half of posts that contained antisemitism. At this threshold the model also achieved a precision score of 72%, meaning that when the model flagged a post as containing an antisemitic conspiracy theory, the model was correct 7 out of 10 times.  

Comments assigned a high likelihood of containing antisemitic hate were then reviewed by researchers, who assessed the posts and identified the top viewed posts that contained the conspiracy. Posts were included in the final dataset where they had been reviewed by two researchers, with both agreeing that the post contained the conspiracy theory that the attack was orchestrated by Jewish people or Israel. 

Researchers then checked each of the 100 top-viewed posts promoting the antisemitic conspiracy theory against X’s Community Notes data and found that 0% of them displayed a Community Note at the time that the research was carried out.