By Kelly-Daniel Oshiogwemue
Nigeria’s formal political campaigns may not yet be the most visible users of generative AI. But their supporters, influencers and informal political networks are already operating in an information ecosystem where synthetic media can be produced, personalised and distributed at very low cost. On May 27, 2026, a video circulated online in Nigeria carrying what appeared to be the voice of President Bola Ahmed Tinubu. The synthetic voice appeared to make inflammatory claims about insecurity, the 2023 election and political opponents. It was not authentic.
The controversy quickly became more complicated. The Presidency initially accused social-media activist Martins Vincent Otse, popularly known as VeryDarkMan, of circulating the material. Subsequent investigations established that the manipulated audio had been superimposed on footage from one of Otse’s original videos. On June 4, presidential spokesperson Bayo Onanuga announced that police had arrested a man identified as Ifechukwu Dennis, whom authorities accused of originating the fake voice. The circumstances of the case and any eventual prosecution remain separate questions.
The episode matters for more than the spectacle of a cloned presidential voice. It exposes a problem that Nigeria’s 2027 election will increasingly have to confront: When artificial intelligence is used to manufacture political content, who is actually responsible for what voters see? The answer may not always be a political party. Nigeria’s formal campaigns are identifiable organisations. Their candidates have names, spokespeople, lawyers and reputations to protect. Their digital communications can, at least in principle, be traced to official structures.
The wider political ecosystem is different. Supporters, influencers, anonymous accounts, partisan communities and informal networks can create and distribute material without necessarily carrying the branding of the campaign they support. Nigeria has seen this before. What is changing is the production technology.
The AI election may be happening outside campaign headquarters
Nigeria’s 2023 election demonstrated how far political communication had moved beyond official party channels. Research by the Centre for Democracy and Development found that online political organisation was often informal and that parties and candidates could operate through unofficial accounts or hired influencers. CDD reported the use of “keyboard/cyber warriors”, sometimes supported by automated bots, to spread disinformation about opponents. Some networks even presented themselves as independent fact-checkers or platforms in an attempt to give political claims greater credibility. The organisation of political influence was therefore already becoming decentralised.
The Overseas Development Institute’s analysis of more than 600,000 tweets during the 2023 campaign similarly found distinct online communities around Peter Obi, Bola Tinubu and Atiku Abubakar. The researchers described the ecosystem as decentralised and noted the emergence of influencers whose importance did not necessarily correspond to formal party structures. That distinction is crucial for understanding AI.
Generative AI does not have to be formally commissioned by a presidential campaign to influence an election. A supporter can generate an image. An influencer can create a synthetic video. A partisan community can produce variations of the same political argument. An anonymous account can distribute the material. A WhatsApp group can give it a second life. The technology therefore lowers the cost of political production without necessarily creating a clear chain of accountability.
Nigeria’s old disinformation problem has acquired a new production tool
There is already evidence of the underlying political behaviour. Africa Check’s review of the 2023 election found that influencers promoted candidates while attacking opponents, sometimes using false information. Theresa Amobi, a senior lecturer at the University of Lagos, told the organisation that politicians hired influencers to spread false information to promote themselves and “demarket” opponents.
CDD’s election assessment similarly documented paid influencers, manipulated audio and video, false impersonation, deliberate mistranslation, automated or controlled networks and cross-platform dissemination. Rumours originating on WhatsApp or Twitter could be picked up by traditional media or influential community figures and then spread further offline.
Generative AI does not create those incentives. It makes them cheaper to act on. A political operator who previously needed a designer, editor, translator or voice actor can increasingly automate parts of that production process. The significance is therefore not simply that AI can make a convincing deepfake. It is that AI can turn political content production into a scalable process.
Idayat Hassan, director of the Centre for Democracy and Development, made this point in a June 2026 analysis for the Centre for International Governance Innovation, arguing that AI is strengthening Nigeria’s existing fake-news economy rather than replacing it. Her description of the country’s information environment is particularly relevant to the 2027 election: informal actors already know how to create and distribute political narratives; AI gives them additional production capacity. That is a more immediate concern than waiting for a perfect presidential deepfake.
The evidence is already appearing
The 2026 election environment has already produced examples of synthetic political media. In May, Africa Check examined an image circulating in Nigeria that appeared to show activist Omoyele Sowore shaking hands with President Tinubu. The image was being used to support a claim that Sowore had aligned himself with the president. Africa Check concluded that the image was AI-generated. In July, another claim attempted to undermine Peter Obi by alleging that he had fabricated photographs from a roundtable in Berlin. Africa Check found that one image appeared to have been enhanced or edited using AI. But independent photographs and video from the European Council on Foreign Relations confirmed that Obi had actually attended the event.
That case illustrates an important complication. AI manipulation does not always create an entirely fictional event. Sometimes it is used to attack the authenticity of a real event. The political effect can therefore come from the uncertainty itself. If supporters can persuade voters that an opponent’s genuine photograph is fake, the objective has partly been achieved even when the underlying event really happened. And the reverse is equally possible: a fabricated image can be used to manufacture evidence for an allegation that supporters of one political camp already believe.
The most dangerous AI content may not look spectacular
The public imagination tends to focus on a flawless deepfake video. But election misinformation rarely needs to be perfect. A genuine photograph can be paired with a false caption. An authentic speech can be given a synthetic voiceover. An old video can be presented as recent. A real politician can be placed into a fabricated scene. A fake screenshot can be designed to resemble an official announcement. The objective may not be to fool every voter. It may simply be to provide enough apparent evidence for a rumour to become believable within a particular political community.
This is where the Nigerian experience with political influencers becomes important. The country’s online political ecosystem already contains communities with strong partisan identities. CDD’s 2023 research found that information was repeatedly moving across platforms and from online spaces into offline conversations. AI gives those communities a new ability: rapidly producing content that fits the narrative they already want to believe.
From propaganda to personalised persuasion
There is an even bigger issue beyond synthetic images. Artificial intelligence can help produce different versions of the same political argument for different audiences. A message about unemployment can be framed around economic anxiety. The same political objective can be reframed around insecurity for another audience. Another version can emphasise religion, ethnicity, regional grievances, youth aspirations or distrust of government. The technology does not need access to a government database to make this possible. Campaigns, advertisers, influencers and online platforms already operate in an environment rich in audience information.
Nigeria’s Data Protection Act recognises political opinions or affiliations as sensitive personal data and establishes rules governing the collection, processing, use and disclosure of personal information. The framework also provides rights relating to access, rectification, erasure, objection and automated decision-making and profiling. The distinction matters. There is no evidence that Nigeria’s major government identity databases have been merged into a political database available to campaigns. Nor should the existence of NIN, voter, banking, telecommunications or other databases be treated as evidence that political campaigns can access them. The policy problem is more precise: What safeguards apply when political actors use whatever personal or audience information they lawfully or unlawfully obtain to build increasingly personalised persuasion systems? That is where AI, data protection and election integrity begin to intersect.
The National Data Protection Commission and INEC appear to recognise the issue. In April 2026, the two institutions established a joint working group to strengthen privacy protections in the electoral system. NDPC said INEC holds the personal data of about 94 million Nigerians and specifically highlighted the need to engage political parties over how they process members’ information. The working group is also expected to address responsible data processing by political parties, civil-society organisations, third parties and vendors. That is significant because it moves the conversation beyond deepfakes. The AI-election problem is also a data-governance problem.
The parties may not be the only actors that matter
There is an important caveat to the argument that the AI election will be driven by supporters rather than parties. The public evidence does not allow us to conclude that Nigeria’s political parties are not using AI. Indeed, the boundary is already becoming blurred. Political support groups are active ahead of the campaign season, while candidates and political organisations are increasingly discussing AI as part of political communication and governance. A recent report, for example, documented the launch of a technology-focused support initiative around President Tinubu, while an opposition political supporter of Peter Obi in Ekiti State called Mr Isaac Fayose said he would deploy technology-driven systems to monitor elections.
What can responsibly be said is narrower: There is still limited public evidence showing that Nigeria’s major presidential campaigns have systematically disclosed large-scale use of generative AI for political persuasion. That lack of transparency is itself worth investigating. If a campaign uses AI internally to draft speeches, analyse sentiment or produce legitimate campaign material, that is very different from a supporter network secretly generating synthetic attacks on opponents. The regulatory question is where the line between the two lies. Who is responsible for the supporter? That may become one of the hardest questions of 2027. Suppose an AI-generated video attacks a presidential candidate. It is posted by an account calling itself an ordinary supporter. The account has no official campaign branding.
But it is repeatedly amplified by hundreds of accounts connected to a political movement. Who created it? Who paid for it? Was it coordinated? Did a campaign know about it? Did anyone instruct its production? Did the campaign benefit from it? These are not merely questions for journalists. They are questions for election observers, regulators and political parties. The International Republican Institute’s August 2026 pre-election assessment makes a similar point from a governance perspective. It recommends that Nigerian political parties and candidates refrain from producing or amplifying false information, manipulated electoral material and AI-generated impersonations. It also recommends that campaigns take responsibility for their official digital communications and promptly correct false content disseminated through their channels. The next step should be extending that accountability conversation to campaign ecosystems, not just official accounts.
INEC lack capacity to police the internet, platforms as part of that architecture
Nigeria’s electoral commission is already treating AI as an electoral issue. INEC created a dedicated Artificial Intelligence Division and has been working on an AI framework for electoral administration. The Commission’s technology work reflects both sides of the issue: AI can improve electoral processes, but it can also be weaponised to manipulate information and undermine trust. In April 2026, INEC also confronted an AI-related impersonation controversy involving material falsely attributed to the Commission’s chairman. The incident demonstrated how easily false digital material can exploit the authority of an electoral institution.
INEC’s official election calendar now lists January 16, 2027 for the presidential and National Assembly elections and February 6 for governorship and State House of Assembly elections. But INEC cannot be expected to determine the truth of every political claim circulating across Nigeria’s internet. It can authenticate its own notices. It can communicate official results. It can correct false claims about voting procedures. It can develop systems for rapid response. It cannot become the referee of every partisan dispute. That requires a broader information-integrity architecture.
The platforms already have rules. Meta says it labels certain AI-generated content, requires disclosure for certain AI-altered political advertisements and maintains political-ad transparency systems. X prohibits deceptive synthetic or manipulated media that can cause serious harm and has a civic-integrity policy covering misleading information that can affect participation in elections. TikTok prohibits paid political advertising and requires clear labelling of realistic AI-generated or edited content. Its civic-integrity rules also prohibit misinformation that could interfere with elections. YouTube prohibits certain election misinformation, including misleading content about voting and technically manipulated material that poses a serious risk of harm.
The existence of these rules is useful. But rules written in California, London or Singapore do not automatically solve a Nigerian election problem. The challenge is enforcement. Can platforms identify a manipulated Hausa-language audio quickly? Can a fake Yoruba political video be escalated rapidly? Can a false INEC announcement be removed before it moves from Facebook to WhatsApp? Can a synthetic video regenerated in five different forms be tracked after the first version is debunked? Those are the operational questions Nigeria should be asking now.
Nigeria can learn from other democracies
Taiwan’s 2024 election offers an important lesson. The country’s information environment faced manipulated videos, rumours and AI-generated material, but the response was not based on one magical detection technology. Fact-checkers, journalists, election authorities, civil-society organisations and public education efforts worked together to contest false narratives. The lesson for Nigeria is not that every deepfake can be detected. It is that resilience depends on institutions that can respond faster than false narratives become entrenched.
Samson Itodo, executive director of Yiaga Africa, made a similar argument in an International IDEA analysis of AI and African elections, saying that electoral authorities and observers need to understand the technology while protecting fairness, reliability, privacy, transparency and inclusion. Gbenga Sesan, executive director of Paradigm Initiative, has also warned that the transformation is fundamentally about scale and speed. Speaking at a March 2026 AI and open-data event, Sesan argued that AI itself is not new; what has changed is the scale and speed at which it can be deployed. That may be the most useful way to think about Nigeria’s 2027 election. The country is not suddenly discovering political manipulation. It is acquiring a technology that can make manipulation faster, cheaper and easier to reproduce.
What Nigeria should do before the campaign peaks
The response does not need to be an attempt to ban AI from politics. That would be unrealistic and potentially harmful. AI can help journalists analyse campaign promises. It can help election authorities communicate with citizens. It can assist civil-society organisations with monitoring. It can improve translation and voter education. It can help citizens understand complicated policy proposals. The goal should instead be provenance, transparency and accountability. Political parties should publish and regularly update their official digital channels. They should establish explicit rules for volunteers, influencers and support groups concerning synthetic media.
Campaigns should disclose AI-generated political advertising and prohibit the impersonation of candidates, election officials and institutions. Political parties should also be expected to respond rapidly when supporters distribute demonstrably false material through channels that are formally or informally connected to their campaigns. Election observers should monitor not only official campaign accounts but also networks of influencers and support groups. NDPC should continue its engagement with political parties over the processing of personal information and election-related data. INEC should maintain a rapid-response mechanism for authenticating election information. And platforms should provide Nigeria-specific escalation channels, including sufficient local-language capacity during the election period.
The first AI election is not necessarily the one with the best deepfake
Nigeria’s 2027 election may not be defined by one spectacular synthetic video. The more consequential change may be quieter. A thousand political messages produced instead of one. Ten versions of the same rumour instead of one. A synthetic image adapted for different communities. An anonymous supporter generating content that a campaign can benefit from without formally owning. A real photograph transformed into “evidence” for a false political claim. A genuine recording dismissed as fake because voters have learned that anything can now be generated. That last possibility may be the most damaging. AI creates a double problem: false evidence becomes easier to manufacture, while genuine evidence becomes easier to deny.
Nigeria’s political information environment already has a history of partisan misinformation, influencer campaigns and weak trust. Academic research based on interviews and focus groups across the country found that social-media content was influencing traditional media while aggravating existing ethnic and religious tensions. Generative AI enters that environment as an accelerator.
The accountability gap is the real election risk
Nigeria should therefore resist the temptation to ask only whether APC, PDP, ADC, NDC or another party will use AI. That is too narrow. The more important question is: What happens when politically motivated citizens can use AI on behalf of their preferred candidates without waiting for those candidates to instruct them? That is much harder to regulate. A campaign headquarters can be identified. A supporter network can be anonymous. A party can issue a correction. A synthetic video may already have travelled through hundreds of WhatsApp groups. A fact-checker can debunk one image. Another version can be generated tomorrow. And a candidate can deny responsibility while still benefiting from the political effect. That is the accountability gap Nigeria needs to close before election day.
The country’s formal campaigns may eventually become sophisticated users of generative AI. Some already appear to be experimenting with technology in their wider political ecosystems. But the evidence available today does not justify assuming that the campaign headquarters will be the centre of the AI election. It may instead happen at the edges, among influencers, supporters, anonymous accounts and/or among political communities that have already learned how to make information travel. The technology has arrived in an electoral environment that already knows how to amplify political narratives.
The question for Nigeria is no longer whether AI will enter the 2027 election. It already has. The harder question is whether the country can establish responsibility before synthetic political content becomes impossible to trace back to the people who created, financed or amplified it. Because in an AI-era election, asking “Who produced this message?” may no longer be enough. Nigeria will also have to ask: Who benefited from it? Who coordinated it? And who is responsible for the political reality it creates?
*Oshiogwemue is a chartered statistician, researcher, entrepreneur, and public affairs analyst.