Tackling threats to election information in the May 2026 elections in England
Key findings on information for voters
- Despite getting information in a range of formats, voters have concerns about the accuracy of information
- More than half of people we surveyed told us they’d seen misinformation ahead of the elections, with 1 in 10 saying they had seen a deepfake
- Ahead of the May 2026 elections we trialled new approaches to protecting voters’ access to accurate and trustworthy election information, based on the kinds of interference we have seen in elections abroad
- Continued collaboration between governments and other organisations is needed to address the developing threats of online misinformation and disinformation, coordinated influence attempts and campaigns, harassment, intimidation and abuse
- The UK government should update electoral law and online platform duties so the platforms can respond to misinformation, disinformation and manipulated media
Voters have concerns about the reliability of election information
Candidates and parties share information with voters in a range of ways
At the 2026 elections, the most common ways people reported seeing information on parties and candidates were:
- a leaflet or flyer from a candidate or political party (63%)
- social media (26%)
- from friends, family or carers via word of mouth (16%)
Older people were more likely than younger people to have seen a leaflet or flyer. Younger people were more likely to say that they had seen information on social media or received information by word of mouth from friends, family and carers.
Candidates also told us that they used campaign leaflets (94% of candidates) and door-to-door canvassing (90%). Just over two thirds (68%) spent time campaigning on social media.
Voters have concerns about the accuracy of election and campaign information
We asked people how much of a problem they thought a range of issues were in England. Alongside low turnout (58%), the issues that people were most concerned about related to election information, these were:
- perceived bias in the media (62%)
- political misinformation or disinformation (57%)
- foreign influence on UK election results through, for example, campaigning activity or disinformation (46%)
- inadequate control of political activity on social media (44%)
This is consistent with what people told us after the May 2025 elections.
Our research found that over half of people said that they had seen misinformation about either party policies (44%), individual candidates (33%) or how elections work (20%). They were more likely to have seen misinformation online than offline.
Most people (65%) who told us they saw misleading or inaccurate information online said that they ignored it. Those who took action said that their most common responses were to fact-check the information or to report it, either to the platform on which it was posted or to someone else.
When asked if they had seen deepfake photos, videos or audio clips in relation to the May 2026 elections, 1 in 10 people (10%) told us that they had, 70% said that they had not, and 20% didn’t know or weren’t sure.
The large proportion of people who said they did not know or were not sure they had seen any deepfake materials suggests that recognising deepfake content may be difficult for some people.
Testing new approaches to protect election information
False or misleading information spreads quickly during elections, especially online. AI tools have made creating convincing fake images, audio and video faster, cheaper and more accessible than ever.
While no significant impact on a UK election has yet been observed, examples from elections in other countries have shown how AI-generated or manipulated content (sometimes referred to as ‘deepfakes’) can pose risks to election security and wider confidence.
Ahead of the May 2026 elections, we commissioned a six-month project with the Accelerated Capability Environment (ACE), a cross-government innovation capability within the Home Office, to detect political deepfakes and understand the threat they pose to voters.
The pilot helped us understand practical ways to monitor, assess and respond to the risks deepfakes pose. It focused on how bringing together cutting-edge technologies and collaborative ways of working could support swift decision-making when deepfakes that posed a risk to the integrity of the elections – or to candidates – were identified.
Our pilot gave reassurance about the integrity of the 2026 elections
During the pilot, our tools automatically processed approximately 3 million online posts. A sub-section of these were reviewed by a person and 219 political deepfakes were identified from 60 unique accounts. The posts were viewed 9.43 million times with 403,000 likes, shares or comments.
However, 15 of these online posts met our threshold of being a risk to electoral integrity. These included false depictions of candidate campaign activities, false candidate announcements and misleading information about polling day.
There was however no evidence detected of deliberate use of highly convincing political deepfakes to target or interfere with the May 2026 elections. The material reviewed suggested that while there is a high volume of AI generated content on social media, it often lacks sophistication.
This does not mean that future risks of political deepfakes are low. The pilot has helped to give reassurance about these specific elections and the technical capabilities to detect deepfakes. But even if political deepfakes are rare they could have a significant impact on elections, as we have seen internationally. We intend to continue this work, to inform voters, candidates and parties.
Social media platforms responded inconsistently to harmful election content
In a small number of cases where content raised serious concerns, we requested removal from social media platforms and reported suspected breaches of the law to the police.
Where we reported material to social media companies, we had varied responses. In some cases, they did act. But other responses fell short of expectations. For example, AI manipulated content reported to X on polling day which made false claims about threats to candidate safety at polling booths was not removed despite appearing to breach the platform’s own guidelines. This material concerned us as it appeared to incite aggression against some parties (in this case Reform UK) and deliberately aggravate community tension with a false allegation.
Harm matters more than manipulation
Our experience was that how, when and where a deepfake is published often matters more than how convincing it is. Content is not harmful just because it is AI generated. The risk depends on how it is received by those viewing it.
Deepfakes may be used for misinformation or disinformation, but they may also appear in satire or legitimate political campaigning. Their significance therefore depends on more than whether a piece of content is AI generated.
This highlights the importance of human judgement. While the outputs of deepfake detection tools are helpful, they cannot be treated as definitive conclusions or determine specific responses. Decisions about the potential threat that a deepfake poses to electoral integrity must therefore rest on human decision making.
This introduces organisational challenges for deepfake detection efforts. Conducting human-led analysis of individual posts requires significant resources and specialist knowledge. It also limits the volume of material that can be reviewed.
The pilot showed the need for a stronger cross-agency response to election information threats
Beyond identifying deepfakes, the pilot created a proof of concept of a system to monitor significant volumes of social media, detect deepfakes, and store them within a secure environment.
The pilot also developed approaches to assess the risks deepfake content pose to electoral integrity and tested ways to securely share information with other partners such as police. This capability could now be scaled and embedded for future elections.
But the pilot also identified significant gaps. Despite high levels of public and parliamentary concern about political deepfakes, they largely fall within a regulatory grey zone. Unless they meet specific conditions, for example depicting intimate image abuse, they are unlikely to breach the law.
In practice, decisions about whether political deepfakes are acceptable or what constitutes an electoral risk currently largely rests with the social media platforms. Our experience of engagement with some of the platforms highlights the unsatisfactory nature of this approach.
Perhaps most importantly, the pilot demonstrated that political deepfakes are just one part of a much wider information threat landscape. They fall within a broader context of online misinformation and disinformation, coordinated influence attempts, harassment, intimidation and abuse, all of which can undermine electoral participation and integrity.
Responding to deepfake-related threats therefore cannot be done in isolation and requires a coordinated response from governments and other organisations that responds to the challenges of online information threats more broadly.
Preparing for future election misinformation and disinformation risks
Our pilot has highlighted that the current policy and regulatory framework for protecting election information in the UK can’t always respond quickly enough to the threat of misinformation and disinformation. It also showed that deepfake-related risks cut across electoral regulation, policing, online safety and public communications and do not fall neatly within a single organisation’s remit.
Collaborative efforts are needed to tackle voter concerns, improve the systems that protect election information and make sure voters have access to accurate and trustworthy information.
Governments, regulators, social media and digital service providers and campaigners need to work together to make sure voters have access to accurate election information and develop an effective response to misinformation and disinformation.
We already work with a range of partners as a member of the UK government’s Defending Democracy Taskforce and Joint Election Security Preparedness Unit. We also coordinate work with other regulators ahead of elections, including Ofcom, the Office for Statistics Regulation, the Charity Commission and the Advertising Standards Authority.
We will continue to provide accurate and impartial election information for voters and campaigners, so they can confidently take part in elections. We will work with online platforms and services and governments to ensure that users can access this information easily during election periods.
Our voter education work with young people also helps them develop the skills to navigate the information environment. This includes identifying misinformation and disinformation, understanding how social media and algorithms shape the information they see, and evaluating different perspectives and sources of evidence.
Recommendation 7: The UK’s governments should consider how the UK can develop capabilities needed to protect the election information environment
Our deepfake detection pilot was one element of a range of monitoring and assessment activities aimed at protecting the election information environment ahead of the May 2026 elections. We will continue to monitor election-related social media mis and disinformation, including deepfakes, to address the threat to the integrity of elections and our democracy. However, it is likely to be more efficient and represent better value for money for the UK to develop further state-level capability, targeted at defending democracy, shared with interested governmental and other bodies, and able to operate at scale.
Building on our experience and findings, the Defending Democracy Task Force should work with all of the UK’s governments to consider how to ensure the UK can continue to develop the technical and operational capabilities needed to protect the election information environment from mis and dis-information. The recent independent report into foreign interference by Philip Rycroft recommended that dealing with hostile state foreign interference should be a higher priority for the government, and that a permanent secretary-level lead be appointed to co-ordinate the response to threats to our democracy. The work across the UK governments should be co-ordinated by them.
This should include how to develop a joined-up approach to identifying and assessing potentially harmful election-related information, including deepfakes and other types of social media and information threats, which is fully scalable from threats to individual candidates to co-ordinated malign interference in our democracy. It should cover both the technical capabilities required to identify potentially harmful information, and the organisational structures and processes needed to share and rapidly assess it.
The UK’s governments should also consider how to ensure the most appropriate and effective response to election-related mis and dis-information, taking into account the findings from our pilot.
Recommendation 8: Electoral law should be updated to reflect new threats to election information
Much of the UK’s electoral law was drafted before the emergence of digital campaigning and new technologies such as generative AI. Governments need to ensure that critical election information is properly protected by laws that can respond to the changing election information environment as our electoral and democratic systems are part of the UK’s critical national infrastructure.
The UK government has previously said that it will consult on how to address the threats posed by ‘digital replicas’. A digital replica is an image, video or recording that copies a person's appearance or voice without their permission. This consultation should include a focus on threats to election information.
For example, the government could consider:
- updating the current legislation that covers false statements about candidates so that it covers the use of deepfakes in an electoral context
- introducing new legislation preventing people from using images, video or audio of candidates or electoral administrators in a way that misleads voters
This could draw on recent legislation in other countries. The Canadian government has recently introduced the 2026 Strong and Free Elections Act to make sure its election law also covers realistic deepfakes.
Recommendation 9: Online services should be required to strengthen policies to limit the spread of misinformation and disinformation
Many large social media platforms and online information services that operate in the UK have content and usage policies aimed at making sure users only get accurate information on elections.
But it is clear that many are not responding quickly enough – if at all – to the spread of false information about elections. Our experience of escalating threats that were identified through our pilot to online service providers indicates that some were more willing than others to remove content in line with their own policies even where we identified what appeared to be a breach in terms of use.
There are already requirements in the Online Safety Act (OSA) for large platforms to protect content of democratic importance when applying content moderation, but the UK government should go further. It should create a clearer, new overarching duty on platforms to cover a wider range of risks to elections, to ensure they take action and protect legitimate political debate, particularly during critical election periods.
This should include a more specific and stringent requirement for platforms to develop fit-for-purpose policies to protect election information, and they should be held accountable by Ofcom for implementing them effectively. This would set clearer expectations on platforms and give the regulator a stronger basis for enforcement where needed.
The government should also amend the OSA to class critical election offences that protect voters and campaigners as ‘priority offences’. This could include new offences that we have recommended to protect voters from misleading deepfakes or other manipulated media. This would mean that platforms would be legally required to actively manage risks to election information and protect legitimate political debate.