Google's AI Watermark Tool Faces Nigeria Election Test

A group of women reading election lists posted on a wall outdoors during the day. Photo: Fatima Yusuf/Pexels

In early July 2025, Nigerian social media platforms saw the spread of a digitally created picture allegedly depicting Peter Obi engaged in conversation with Sani Abacha over tea or coffee. Since Abacha passed away in 1998, Obi immediately dismissed the claim of any meeting between them. Verification teams later confirmed the image was entirely fabricated. During the same period, a viral TikTok clip falsely reported that President Bola Tinubu had fallen gravely ill from poisoning, following the death of Muhammadu Buhari.

Additionally, a manipulated video from 2024—constructed from 2023 footage paired with an altered voice track—falsely suggested Tinubu had pledged to sell fuel at ₦100 per liter, accumulating roughly 89,000 views before being disproven. Leading up to the 2023 elections, a doctored audio recording falsely suggested Atiku Abubakar, Ifeanyi Okowa, and Aminu Tambuwal were conspiring to rig the vote. Investigators determined the clip was pieced together from unrelated segments of their past speeches.

Google Introduces a New Verification System

Google has recently made its SynthID detection system accessible to the public as a standalone platform. This free service evaluates whether digital media, images, videos, or audio, originates from artificial intelligence. The technology relies on research from Google’s DeepMind division, developed since 2023, and functions by identifying hidden digital markers embedded during AI generation.

According to Google, over 180 billion images and videos, along with audio equivalent to 240,000 years of speech, now carry these watermarks. The detector is now integrated into Google Search, the Gemini app, and Chrome, processing more than one million verification requests daily. Collaborating entities include OpenAI, NVIDIA, and Kakao, with Apple expected to participate soon. While the tool appears promising for safeguarding election integrity, its effectiveness in Nigeria presents significant challenges.

The SynthID system can only detect AI-generated content stamped with Google’s watermark or those from its partner organizations. Media produced using external tools, such as open-source generators, budget-friendly mobile applications, or lesser-known AI platforms, remains undetected. This leaves room for politically motivated campaigns, foreign interference operations, or even individual creators in Lagos to distribute manipulated content that evades detection. Furthermore, the tool operates exclusively in English, limiting its accessibility in a nation where political discourse thrives in Hausa, Yoruba, Igbo, Pidgin, and numerous other languages.

Nigeria is not entirely defenseless, but existing measures remain inconsistent and unevenly distributed. The Independent National Electoral Commission (INEC) established an Artificial Intelligence Division in May 2025 to address election-related risks, including disinformation and synthetic media, while enhancing data-driven decision-making and voter services. Meanwhile, the Center for Collaborative Investigative Journalism launched ElectionWatch, an AI-driven platform tracking misinformation on TikTok and Telegram, supported by the Google News Initiative via JournalismAI. FactCheckAfrica’s MyAIFactChecker provides rapid credibility assessments in Hausa, Yoruba, and Swahili, an important step toward language-inclusive verification tools. However, it currently cannot authenticate deepfake audio, images, or video, though updates are planned.

Google’s Tool and Nigeria’s Election Challenges

Google’s detector provides a method to verify the origin of AI content, allowing readers to check what they consume before sharing. However, for the tool to function as election infrastructure in Nigeria, it must operate in the languages Nigerian communities speak. The country is now 18 months away from its next major election cycle, and AI-generated political content is well documented, recurring, and becoming increasingly sophisticated.

Why the 2027 Election Cycle Differs

Since 2015, Nigerian elections have consistently faced digital interference, but the scale of manipulation has grown far more accessible. Crafting a convincing audio deepfake once required specialized expertise; today, it can be produced with a free AI account in about 20 minutes. In June 2025, following a Benue State attack that killed over 100 people, a video falsely showing Nigerian Army personnel escorting livestock circulated online. PRNigeria’s analysis confirmed it as an AI-generated deepfake intended to provoke conflict.

Separately, another video claiming to depict a Benue vigilante-herder confrontation was revealed as archival footage from a 2022 Kinshasa gang clash in the Democratic Republic of Congo. Deepfake videos featuring Donald Trump in relation to Nigeria, including fabricated endorsements of Nigerian candidates, discussions about Biafra, and criticism of President Tinubu, have also spread widely. A Dubawa study found that around 75% of Nigerians encounter misleading online content, with 68% admitting they have trusted and acted on false information. Nigeria’s ranking of 112th out of 180 in the 2024 World Press Freedom Index further complicates efforts by fact-checkers to combat disinformation.

Preparing for February 2027

The SynthID Detector represents meaningful progress by enabling users to verify AI-generated media before sharing it. However, for Nigeria to fully leverage it as an election safeguard, the tool must support the country’s dominant languages. Nigeria’s Electoral Act prohibits false or deceptive candidate statements, with penalties including fines and imprisonment, while the Cybercrimes Act has been applied to manipulated content, though critics argue it has also suppressed legitimate discourse. Currently, no legal requirement exists for campaigns to disclose AI use in their materials.