This report presents the consolidated findings of Build Up’s social media monitoring and analysis conducted between 1 August 2025 and 28 February 2026, covering Bangladesh’s general elections and July Charter referendum. Drawing together insights from the baseline study, continuation analysis, and three snapshot reports, it provides an integrated account of how digital narratives and harmful behaviours evolved across the full electoral cycle. 

Bangladesh’s 2026 elections took place within a deeply contested political landscape shaped by the July 2024 uprising, the resignation of former Prime Minister Sheikh Hasina, and the establishment of an interim government tasked with institutional reform and electoral oversight. In this context, social media platforms became central arenas where political actors, media institutions, influencers, and citizens interpreted political developments, contested legitimacy, mobilised support, and advanced competing visions of the country’s democratic future. 

Drawing on data from Facebook, TikTok, and X, the analysis tracks online political discourse with a focus on harmful digital content and behaviours, including mis/disinformation, hate speech, incitement to violence, and tech-facilitated gender-based violence, alongside constructive narratives related to de-escalation, human rights, and social cohesion. Across approximately 499k posts classified during this period, the data demonstrate that digital harms were not isolated incidents but recurring, structural features of online political discourse throughout this critical democratic transition. 

The fourteen key findings and six recommendations that follow are intended as a practical resource for civil society organisations, media actors, policymakers, and international partners working to safeguard civic space, information integrity, and democratic governance in Bangladesh’s evolving information ecosystem.

Key Findings

• Key Finding 1 | Hate speech was the most prominent digital harm observed, accounting for 8% of all classified posts, with polarising and dehumanising language so normalised within political discourse that it functioned less as an aberration and more as a default register for political contestation. Two dominant patterns were identified: politicised hate speech deploying terms such as “terrorist” and রাজাকার to delegitimise political opponents, and religious and communal hate speech targeting minority communities — particularly visible around Hindu festivals and debates on religious identity.
• Key Finding 2 | 35.6k posts were identified as mis/disinformation, with misleading narratives consistently amplified by high-profile accounts before verification could occur — exploiting the information vacuum that surrounds electoral processes to shape perceptions of institutional legitimacy. Three recurring dynamics were observed: speculation around electoral procedures and outcomes, the recycling of political rumours and conspiracy narratives across reporting periods, and amplification through influential accounts whose reach allowed unverified claims to circulate widely before correction was possible.
• Key Finding 3 | 28.6k posts invoked foreign interference narratives, primarily implicating India, functioning less as evidence-based claims and more as a rhetorical tool to delegitimise political actors and frame domestic politics through a lens of external threat and compromised sovereignty. These narratives drew on historical grievances and regional power dynamics, with India’s continued hosting of former Prime Minister Sheikh Hasina a recurring flashpoint; posts referencing western influence, international organisations, and “foreign agents” were also widely circulated.
• Key Finding 4 | Although the smallest harm category at 14.2k posts, incitement to violence spiked sharply around key political events – most notably the assassination of Sharif Osman Hadi – demonstrating how moments of political shock can rapidly translate online anger into calls for retaliatory action. Content ranged from explicit calls for confrontation and targeted threats toward political actors to emotion-driven responses that framed mob violence as a legitimate political response, with spikes consistently tied to moments of heightened offline tension.
• Key Finding 5 | The most volumetrically significant narratives were harmful – centring on claims of violence, delegitimisation of the interim government, and allegations that the July revolution was propaganda – yet the single largest narrative category across the entire monitoring period was constructive, with calls for peace and national unity outpacing any individual harmful theme. This coexistence reflects a genuinely contested information space in which civil society and public figures mounted a sustained counter-effort to stabilise discourse, even as organised political actors drove polarisation.
• Key Finding 6 | With over 16.6k posts and 5 million comments, claims of violence and massacre were among the most volumetrically significant harmful narratives, amplifying fear and perceptions of systemic instability during moments of electoral uncertainty, often without verified basis. These posts did not necessarily reflect the on-the-ground situation but contributed to shaping a perception of insecurity within the digital space, particularly in the lead-up to election announcements.
• Key Finding 7 | 15.9k posts contested the legitimacy of the July 2024 revolution and its aftermath, framing competing interpretations as deliberate deception, and deepening a zero-sum information environment where consensus on basic political facts became increasingly difficult to establish. By dismissing alternative perspectives as coordinated attempts to mislead the public, these posts blurred the line between legitimate political critique and disinformation, reinforcing mistrust across political and social groups.
• Key Finding 8 | Narratives questioning the legitimacy and neutrality of the interim government were consistent across the full monitoring period and spanned the political spectrum, eroding institutional trust and reinforcing scepticism about the credibility of the electoral process at a critical democratic moment. While driven primarily by politically aligned actors rather than broad-based public sentiment, these narratives shaped perceptions of governance throughout the electoral cycle and were particularly prominent around key administrative decisions.
• Key Finding 9 | Accusations of disinformation were themselves weaponised as a political tool, with 13.6k posts using claims of “fake news” and propaganda to delegitimise opposing viewpoints, compounding rather than clarifying the contested information environment. In many cases, accusations of disinformation reinforced existing political divisions rather than improving information quality, making it increasingly difficult to distinguish verified information from narrative-driven claims.
• Key Finding 10 | The largest single narrative category across the entire monitoring period was constructive — 18.3k posts calling for calm, tolerance, and national unity, indicating that civil society and public figures mounted a sustained and meaningful counter-effort to stabilise discourse, particularly during moments of heightened tension. Calls for peace were often reactive, emerging in direct response to spikes in harmful content, and served both as a stabilising force and as an indicator of the perceived fragility of the broader environment.
• Key Finding 11 | TFGBV represented a significant and persistent dimension of electoral discourse, with 4k posts documenting explicit misogyny, gendered abuse, and AI-manipulated content disproportionately targeting women in public life, revealing how digital platforms became active sites for reinforcing gender inequality and suppressing women’s political participation. Abuse ranged from sexually explicit insults and reputational attacks to AI-generated content designed to humiliate and discredit women candidates, journalists, and activists, intensifying during key political events and reinforced by patriarchal narratives linked to certain political and religious actors.
• Key Finding 12 | Despite sustained digital polarisation throughout the electoral period, the 12 February elections were largely peaceful and procedurally orderly, demonstrating that online discourse in this context functioned less as a reflection of public sentiment and more as an amplifier of perceived instability, shaped by a small number of highly active, politically motivated actors. This online-offline divergence does not diminish the risks posed by harmful narratives: sustained exposure to polarised and misleading content contributes to the erosion of institutional trust, normalisation of hostile discourse, and longer-term societal fragmentation.
• Key Finding 13 | Online discourse was disproportionately driven by political parties, affiliated media cells, and high-visibility commentators rather than ordinary citizens – a structural concentration of influence that explains both the intensity of digital polarisation and its limited correspondence with broader public behaviour at the ballot box. A relatively small but vocal group of organised or semi-organised actors shaped dominant narratives through emotionally charged messaging, strategic framing, and, in some cases, AI-generated content, with their reach amplified through platform algorithms.
• Key Finding 14 | UK Bangladeshi diaspora discourse was deeply polarised and emotionally charged, dominated by pro-Awami League sentiment that stood in marked contrast to broader opinion within Bangladesh, with flashpoints including the Tulip Siddiq affair, Tarique Rahman’s return, and concerns over Hindu minority rights generating significant cross-platform engagement. The diaspora context illustrates how Bangladesh’s information environment extends beyond its borders, with UK-based actors actively shaping and amplifying narratives that fed back into domestic political discourse.

Recommendations

Recommendation 1 | Strengthen platform accountability and content moderation. Facebook, which accounts for over 70% of social media usage in Bangladesh, must be held to higher and more consistent standards of content moderation during politically sensitive periods; TikTok’s exceptionally high engagement ratios and documented use of AI-generated content require platform-specific transparency reporting in Bengali and increased moderation capacity.

Recommendation 2 | Strengthen BTRC’s regulatory role and formalise coordination with civil society and experts. The Bangladesh Telecommunication Regulatory Commission must evolve from a reactive regulator into a proactive, rights-respecting body capable of addressing digital harms at scale, with structured mechanisms for civil society input into regulatory design and review.

Recommendation 3 | Enact and reform laws to effectively address hate speech and disinformation. Bangladesh’s existing legal framework has not kept pace with the scale, speed, or sophistication of digital harms documented in this report; a modern legal architecture is needed that addresses organised hate speech, AI-generated content, and coordinated disinformation while remaining rights-respecting and proportionate.

Recommendation 4 | Strengthen protections against tech-facilitated gender-based violence. Platforms must enforce content policies specifically covering gendered abuse including AI-manipulated imagery; legal frameworks must provide meaningful remedies for victims; and civil society organisations working on digital safety must be resourced to provide reporting mechanisms and direct support — with TFGBV protections integrated into electoral integrity planning rather than treated as a secondary concern.

Recommendation 5 | Promote media literacy and digital resilience among the public. Given the significant divergence between online narratives and offline realities, digital literacy programmes should be embedded in educational curricula and delivered through community networks, with particular attention to younger audiences and the spread of AI-generated and manipulated content.

Recommendation 6 | Engage electoral institutions in digital harm response. The Election Commission should incorporate digital harm monitoring into its electoral integrity frameworks, formalise procedures for handling online complaints related to mis/disinformation, candidate harassment, and incitement to violence, and ensure that findings from monitoring reports feed directly into electoral review and future regulatory reform.