Synthetic Politics and Counter-Framing: AI-Generated Political Misinformation and Fact-Checking in Bangladesh’s Election-Cycle Information Environment, 2025–2026

Description

This supplementary dataset accompanies the article “Synthetic Politics and Counter-Framing: AI-Generated Political Misinformation and Fact-Checking in Bangladesh’s Election-Cycle Information Environment, 2025–2026.” It contains 122 item-level fact-check records and 13 aggregate monitoring or review sources. The file documents claim descriptors, formats, narrative frames, audit status, data-quality notes, key aggregate figures and live retrieval links for transparent verification and reuse.

Steps to reproduce

The dataset was developed through qualitative framing analysis supported by descriptive content coding. The item-level corpus covers 1 January 2025 to 20 February 2026. Relevant fact-checks were identified by manually browsing the AI, deepfake and political sections of Rumor Scanner, Dismislab, FactWatch, BOOM and Ajker Patrika and by conducting Bengali and English keyword searches. Search terms included AI, artificial intelligence, deepfake, generated image, generated video, voice clone, কৃত্রিম বুদ্ধিমত্তা, এআই, ডিপফেক, এআই ছবি and এআই ভিডিও. These were combined with terms relating to elections, campaigns, nominations, referendums, political parties, public institutions and major political actors. A record was included when the fact-check explicitly identified an AI-generated or AI-edited visual or audio element and the claim concerned a political actor, party, government or security institution, election process, campaign, referendum or public political controversy. The unit of analysis was one unique verified misinformation claim, rather than each social-media post or fact-check article. When multiple organisations examined the same claim, the report containing the clearest accessible verification evidence was retained as the primary record, while additional reports were treated as corroborating sources. The final register contains 122 item-level records: 110 from Rumor Scanner, six from Dismislab, four from FactWatch, one from BOOM and one from Ajker Patrika. Each record was coded for source, year, English claim descriptor, synthetic-media format, dominant narrative frame, audit status and retrieval URL. Eleven records requiring clarification, exclusion, unit review or duplicate treatment are separately identified. Thirteen aggregate monitoring reports provide contextual statistics and are not counted as additional misinformation cases. The workflow was manual and document-based; no laboratory instruments or reagents were used. Initial screening totals were not recorded, so the dataset should be treated as a purposive register of verified cases rather than a systematic review or an estimate of all AI-generated political misinformation circulating in Bangladesh.

Institutions

• Comilla University

  Chittagong, Comilla

Categories

Funders

• International Association for Media and Communication Research

  Île-de-France, Paris

  Grant ID: Journalism and Fact-Checking

Licence