The dissemination of deepfake videos on social media has the potential to generate disinformation and influence public perception. This study aims to analyze a deepfake video featuring a likeness of Prabowo Subianto on social media, utilizing Norman Fairclough’s Critical Discourse Analysis (CDA) model. A qualitative approach employing the CDA method was adopted. Research data consisted of speech excerpts, transcripts, screenshots, and visual and audio elements from the deepfake video including a visible WhatsApp number. Primary data were sourced from the deepfake video as reported by the Detik.com and Liputan6 SCTV TikTok accounts, while secondary data were derived from original speeches, online news reports, and relevant scholarly articles. Data collection involved digital observation and documentation, and analysis was conducted using Fairclough’s model, encompassing the dimensions of text, discourse practice, and sociocultural practice. The findings indicate that the deepfake video leveraged persuasion strategies, visual and audio constructions mimicking reality, and the symbolic legitimacy of the presidential figure to build audience trust. Furthermore, dissemination via social media and the public's vulnerability regarding digital literacy exacerbated the potential for disinformation. The study concludes that deepfakes constitute a discourse practice capable of shaping perceptions and reinforcing the spread of disinformation within the digital sphere.

Penyebaran video deepfake di media sosial berpotensi menimbulkan disinformasi dan memengaruhi persepsi masyarakat. Penelitian ini bertujuan untuk menganalisis video deepfake yang menyerupai Prabowo Subianto di media sosial menggunakan model Analisis Wacana Kritis (AWK) Norman Fairclough. Penelitian ini menggunakan pendekatan kualitatif dengan metode Analisis Wacana Kritis. Data penelitian berupa potongan ujaran, transkrip, tangkapan layar, serta unsur visual dan audio dalam video deepfake, termasuk tampilan nomor WhatsApp yang muncul secara visual. Sumber data primer diperoleh dari video deepfake yang diberitakan melalui akun TikTok Detik.com dan Liputan6 SCTV, sedangkan sumber data sekunder diperoleh dari video pidato asli, berita daring, serta artikel ilmiah terkait. Teknik pengumpulan data dilakukan melalui observasi digital dan dokumentasi, sedangkan analisis data menggunakan model Norman Fairclough yang meliputi dimensi teks, praktik wacana, dan praktik sosial budaya. Hasil penelitian menunjukkan bahwa video deepfake memanfaatkan strategi persuasi, konstruksi visual dan audio yang menyerupai realitas, serta legitimasi simbolik figur presiden untuk membangun kepercayaan audiens. Selain itu, penyebaran melalui media sosial dan kerentanan literasi digital masyarakat turut memperkuat potensi disinformasi. Penelitian ini menyimpulkan bahwa deepfake merupakan praktik wacana yang mampu membentuk persepsi dan memperkuat penyebaran disinformasi di ruang digital.