A surge of AI-generated misinformation targeted voters in the Makerfield by-election this month, researchers at the Social Market Foundation said.

The findings highlight the growing vulnerability of local electoral contests to coordinated digital deception. By leveraging hyper-local social media groups, these campaigns can bypass traditional news scrutiny to influence voter behavior through fabricated claims.

Data from the report shows that approximately one in six posts shared within local Facebook groups regarding the election were false [1]. This represents about 16.7% of the content shared in those specific digital spaces [1]. The misinformation specifically targeted Andy Burnham, the Labour candidate for the seat.

Researchers identified a 19-fold rise in misinformation targeting Makerfield voters compared with previous elections [2]. This spike included the use of AI-generated fakes designed to undermine the candidacy of Burnham and sway public opinion [2].

The Social Market Foundation said the coordinated nature of these campaigns suggests a deliberate effort to disrupt the democratic process. Because the content was spread through community-led Facebook groups, it often appeared to come from trusted neighbors rather than political operatives.

Local officials and digital watchdogs have expressed concern over the speed at which these fakes proliferate. The use of artificial intelligence allows for the rapid creation of convincing but entirely fraudulent narratives, making it difficult for voters to discern fact from fiction in real time.

One in six posts shared in local Facebook groups regarding the election were false.

The Makerfield case serves as a critical example of 'micro-targeting' where AI is used not just for broad advertising, but to poison the information ecosystem of a specific geographic area. The 19-fold increase in fake content suggests that bad actors are now treating local by-elections as testing grounds for AI-driven psychological operations before applying them to larger national contests.