Online targeted political advertisements

TPAs use tracking technologies (including cookies and AI-powered tracking systems) to collect information from people’s browser history and social media activity. The data is then used to create a profile and draw inferences about people, including their political affiliation and the topics they might be interested in when it comes to voting. Personalised adverts and messages are then created and posted, targeting people based on their profiles. These are posted on websites, social media, TV and streaming platforms etc.

This form of micro-targeting is similar to targeted advertising – that is, using personal data to post adverts that are likely to resonate with people’s interests and beliefs. 

TPAs are used to reach specific audiences and encourage their endorsement of a political party or candidate. Political campaigns in the UK have recently used targeted adverts to reach minority groups and communicate messages on migration, crime and employment. 

TPAs make it easier for political parties and candidates to identify and respond to individuals’ concerns and personal interests and reach people at scale. This could encourage those who are less engaged in politics to vote.

The use of TPAs runs the risk of producing echo chambers – where people’s concerns are played back to them without context or balanced argument, which can contribute to the entrenchment of extremist positions.

Other risks associated with the use of TPA include spreading misinformation, individual profiling and privacy violations. Profiling is the practice of categorising people, often based on harmful stereotypes, and then making predictions about their behaviour. The Cambridge Analytica scandal in 2018 exemplifies these issues.

In 2023, Meta decided to ban political advertisers from using generative AI – a specific type of AI that generates outputs (text, images or video) in response to a prompt. In 2021 Facebook banned TPA based on sexual orientation, race and religion. However, the persistent lack of transparency around who is being targeted, how and by whom, and the lack of necessary guardrails, means it is difficult to assess risks around privacy, misinformation and political manipulation.