14 November 2025

---

For more than a decade, researchers have tracked the role of hostile states in fostering divisive rhetoric online targeting countries including the United Kingdom. However, an analysis by ISD in collaboration with The Bureau of Investigative Journalism (TBIJ) identified a commercial operation spreading anti-migrant and anti-government messaging, seemingly purely for profit. 

Together, we identified a network of more than 100 Facebook pages and groups that form part of a commercial social media campaign run from Sri Lanka. In total, these accounts have more than 1.6 million followers, with the most popular post receiving almost 23,000 shares. Posts from the network include (often AI-generated) anti-migrant content, at least some of which contravene Facebook’s policy on hateful speech. 

This analysis points to the online harms posed by commercially motivated actors alongside those driven by politics: although not politically motivated, their divisive content continues to receive high levels of engagement, fostering hatred and polarisation online. It also reflects the ease with which AI-generated content can help such actors quickly generate high levels of traction, as well as continued failings in platform enforcement. 

Key Findings: 

• ISD identified 128 Facebook entities (89 pages and 39 groups) linked to a coordinated network based in Sri Lanka. The overwhelming majority of them pushed anti-migrant content tailored to the current UK political context.
• The 89 pages collectively attracted more than 1.47 million followers, while the 39 groups gathered 150,000 members, reaching a combined audience of 1.6 million.
• Most pages and groups were created or repurposed between April and September 2025. Many older pages were rebranded to focus on UK political or anti-migrant themes over the last three months.
• 56 pages ran advertisements on Meta, including political ads that appeared to contravene platform policies.
• The network’s content achieved significant visibility: the most viewed post, a video from “Unite the Kingdom” rally, received 8.3 million views. The 10 most viewed posts collectively received 45 million views.
• The published content included hate speech directed at migrants, calls for anti-government/anti-migration protests and mass deportations, demands for new elections, and a range of false claims: these included a video alleging that Prime Minister Keir Starmer had resigned, an AI-generated clip depicting him kissing Labour Party donor Waheed Alli, and debunked claims about counter protesters being arrested at the Unite the Kingdom rally.
• Despite the overtly political tone, evidence suggests the operation was driven by commercial interests rather than ideology. Individuals behind the network offered paid online training on how to monetise political content and use AI-generated imagery to target audiences more effectively. Nevertheless, the anti-migrant content shared by this network is likely to have political consequences. 

Methodology 

A joint investigation by ISD with TBIJ identified a small cluster of Facebook pages displaying signs of coordinated activity. For example, this includes posts related to the Raise the Colours campaign, which encourages public display of the Union Jack and St. George’s flags. The initial cluster was identified when monitoring hate speech against migrants in the UK. Using a snowball sampling method, analysts then expanded the sample to uncover additional pages and groups from within the same network. 

Indicators that suggest the accounts are part of a network include: 

Tactics, Techniques, and Procedures (TTPs): 

• Cross-interaction between pages (mutual follows, coordinated commenting, shared administration of groups and engagement), 
• Use of Facebook check-ins by pages, seemingly to create the appearance of local presence. 

Content-related indicators: 

• Identical or near-identical advertising copy across pages,  
• Repeated use of generic patriotic captions and hashtags (e.g. proud to be British, we love the UK, we are United Kingdom), 
• Consistent use of AI-generated content featuring anti-migrant and anti-government messaging, 
• Characteristic linguistic errors across pages (notably misuse of the verb “to be”) and typos (e.g. “fallow” instead of “follow”), 
• Many pages categorised as “gaming video creators”,  
• Use of fabricated UK addresses in the page description, mostly connected to unrelated businesses,  
• The inclusion of generically British names and patriotic-themed introductions.

Manual verification confirmed 89 pages and 39 groups. However, as new accounts are consistently added to the network, this is not an exhaustive list but a snapshot of the network at the time of data collection (completed on 8 November 2025).  

ISD analysts then used the social media monitoring tool Brandwatch to examine the network’s level of engagement and most popular posts. 

Further analysis of early-stage connections led analysts to identify an individual in Sri Lanka who appears to run an “online academy” on monetising Meta content via Facebook. The academy’s videos featured specific pages belonging to the network, confirming the link between the academy, and the identified pages and groups.  

The network 

The network, consisting of 89 Facebook pages and 39 groups, overwhelmingly used British patriotic slogans and imagery to attract audiences. Despite frequent grammatical errors and poorly phrased English, the network successfully built a substantial following over the course of the past year, with a total of more than 1.6 million followers and members.   

The pages were mostly created between April and September 2025, though ISD and TBIJ identified older pages that showed an abrupt shift in their content. For example, one page representing a Mexican church suddenly pivoted to AI-generated anti-migrant content last year.

More than 90 percent of the pages and groups in the network featured content related to recent migration discourse in the UK. Overwhelmingly, content related to migration was highly negative. However, some pages occasionally shared images and other content which was supportive of immigration, further suggesting the commercial rather than political nature.  

The network’s most popular post featured images from the anti-migrant ‘Unite the Kingdom’ rally on 13 September 2025, organised by far-right activist Tommy Robinson and reportedly attended by up to 150,000 people. The post was accompanied by the caption “Britain is awake”, and received 8.3 million views and 23,000 shares.  

On the same day, the Sri Lankan operator posted on his personal Facebook page asking collaborators managing the “UK pages” to report on their reach, remarking: “10k likes in just one–two hours.” 

The network’s content often borders on anti-migrant hate, using language such as ‘replacement’ (echoing the white nationalist Great Replacement conspiracy theory) and “invaders” to describe migrants. One of the images, alleging to show the future of Britain, included an AI-generated image of leering dark-skinned Muslim men targeting a blue-eyed white woman on a bus; other posts described Islam as “a cancer”. At least one image showed a banner stating “White lives built Britain”, using a photograph that originated from Patriotic Alternative, a British white nationalist movement. 

There was a strong anti-Labour party sentiment across the network, even among the pages that did not explicitly comment on migration. The fifth most popular post voiced support for Ant Middleton, a former British soldier who has expressed support for the Reform party and participated in the Unite the Kingdom rally. The seventh most popular post by likes asked whether Keir Starmer was the “worst UK prime minister in British history”; it received 85,000 likes and 3,600 shares. 

These networks’ success is evidenced by the high levels of engagement and the fact that their content was shared by accounts belonging to public figures, including two Reform UK politicians. The content received positive engagement from two additional pages and a councillor affiliated with the same party. Two other councillors (one from the Green Party and one from the Labour Party) interacted negatively with the content, with one of them repeatedly accusing the accounts of being operated by bots. This engagement from political figures demonstrates the network’s high visibility within UK political discourse, despite its inauthentic nature. 

Tactics  

Generative AI content 

AI-generated content has become easier to produce rapidly and increasingly realistic, driving its adoption by those wishing to promote misinformation or divisive rhetoric. This phenomenon is evidenced by the Sri Lankan network, which featured both images and videos created using generative AI tools. 

A number of the AI-generated symbols or caricatures were not intended to be realistic. For example, the network posted an image of Santa Claus with Union Jack flags and a caption claiming that the Labour government was looking to “outlaw Christmas”, or an image of Britannia holding a banner which reads “only traitors help invaders”.   

However, many posts featuring lifelike images were later determined to be AI-generated upon closer inspection. These images often focused on marches and protests, constructing images of crowds, sometimes with people holding banners or placards. As in the examples below, these posts featured anti-migrant slogans such as the idea of “taking back Britain” or “stop importing start deporting”. The single-most-popular AI-generated image featured soldiers stood at the Cliffs of Dover, with a caption asking whether users wanted the military to “secure the borders”; it received 123,000 likes. 

Facebook’s policy on labelling AI-generated content cites examples such as “a video that appears realistic of a group of people walking around an outdoor market” as material that should clearly carry an AI-generated label. Yet, ISD’s analysis found no evidence of such labels on any comparable content from this network. This absence points to a significant gap in the enforcement and effectiveness of Facebook’s AI detection and labelling systems. 

Monetisation 

Another hallmark of this network is that despite the highly political nature of the material, the operation appeared to be purely profit-driven. In training videos from the ‘online academy’, the operator demonstrated how to identify a specific target audience and set up a political campaign, explaining that political niches have the highest potential to go viral quickly. The creator alleges that AI-generated reels achieve virality up to 10 times faster than other types of content, and demonstrates how to use ChatGPT to find trending news topics and produce relevant AI-generated content. 

The training, which is run from a verified Facebook page marked as an ‘Educational Consultant’, explains how to make use of the Content Monetisation (CM) tool on Meta “to earn money from different content formats, including longer videos, Reels, and photo and text posts, with earnings based on engagement, views, and plays.” Meta’s CM Tool was released in October 2024 and is currently available for a restricted number of creators that adhere to Meta’s monetisation policies. Although content that depicts or discusses immigration in a polarising or inflammatory manner is listed as a “restricted” category which “may face reduced or restricted monetisation”, this guideline to be poorly enforced. 

Besides violating and evading monetisation policies, the network also successfully bypassed Meta’s policies on political advertising and social issues. Ads identified in Meta’s Ad library that had not been moderated included those calling for a ban on burkas in the UK, an AI-generated image of London Mayor Sadiq Khan and Starmer in prison uniforms with the caption “the best possible outcome for Britain”, an AI-generated image of Winston Churchill aiming a rifle at Starmer, ads referencing Reform UK, and an ad urging users to “stop voting for liars in power.” 

Conclusion 

This investigation demonstrated how economically motivated actors orchestrated an online political campaign, exploiting Meta’s vulnerabilities and circumventing regulatory measures. As revealed by the ‘online academy’ behind the network investigated in this dispatch, political content – particularly polarising themes such as immigration – offers a lucrative market for Meta creators. Limited media literacy and a lack of awareness about such inauthentic networks have enabled this network to infiltrate political discourse in the UK – this is evident in the substantial level of engagement (including from politicians) the network received, despite both clear signs of inauthenticity and policy violations. 

While most concerns about information operations currently focus on foreign state actors, the nature of this network’s content should be equally concerning. The widespread availability of generative AI tools have enabled the rapid creation of compelling, high-quality visuals, helping the network to reach a wide audience. This case study exemplifies a shifting media landscape in which audiences struggle to discern inauthentic content, allowing for bad actors to exploit politically charged issues for economic gain.