South Asia, home to some of the world’s largest democracies, has seen a wave of elections in recent years. India held its massive general election in 2024, Pakistan and Bangladesh voted amid political tensions that same year, and smaller polls followed in places like Sri Lanka. Social media platforms, used by hundreds of millions, carry campaign messages, debates, and news. With the rise of artificial intelligence tools that create realistic videos, audio, and images—known as deepfakes—fears grew that fake content would sway voters, spread hate, or undermine trust in results. Headlines warned of chaos, with experts predicting AI could tip balances in tight races.
This debate matters because fair elections build stable governments and protect rights. If misinformation from AI truly destabilizes them, it threatens democracy in a region with histories of polarization along religious, ethnic, and caste lines. Yet evidence must guide judgment, not alarm. This article examines five circulating claims about AI-generated misinformation’s role in South Asian elections, drawing on fact-checking reports, platform data, and studies up to late 2025. It shows instances of use, but limited proof of widespread destabilization—revealing a tool that amplifies old problems more than creates new crises.
Claim 1: AI Deepfakes and Fake Content Flooded South Asian Elections, Reaching Most Voters and Changing Opinions
Reports highlight deepfakes in India (2024), Pakistan, and Bangladesh, including revived deceased leaders endorsing candidates or manipulated voices spreading false statements. Surveys suggest 75 percent of Indians encountered deepfakes during polls.
Exposure occurred, but scale varied. Meta’s 2024 transparency data shows AI content formed less than 1 percent of election misinformation on its platforms across countries, including India and Pakistan. Fact-checkers like Boom and Alt News debunked hundreds, but most viral fakes were “cheapfakes”—simple edits, not advanced AI. Google’s Project Shakti in India found only 2 percent of reviewed stories involved AI.
Historically, misinformation thrived pre-AI via doctored photos or rumors in diverse, low-literacy settings. AI lowers barriers, but traditional methods dominate. Trade-off: AI enables quick personalization in many languages, reaching regional voters—but detection improves, limiting spread.
Verdict: Misleading. AI content appeared and reached many, but formed a small share of overall misinformation.
Claim 2: AI-Generated Misinformation Directly Caused Violence or Major Voter Shifts in Elections
Some point to inflammatory deepfakes fueling tensions, like altered clips stoking communal fears in India or opposition smears in Bangladesh.
No clear causal link emerges. Studies on 2024 elections (Rest of World tracker, Munich Security Conference) find AI used more for engagement—translated speeches, memes—than deception. In Pakistan, Imran Khan’s party used AI voice for jailed leader’s messages, boosting support without proven violence. India’s polls saw communal narratives, but rooted in longstanding divides, not solely AI.
Logically, voters weigh multiple factors—economy, caste, local issues. Single fakes rarely sway masses without fertile ground. Contradiction: AI aids positive outreach, like multilingual campaigns, but risks amplifying bias if unchecked.
Verdict: False. AI contributed to misinformation environment, but no evidence it alone drove violence or decisive shifts.
Claim 3: Political Parties Widely Used AI to Spread Lies, Undermining Election Fairness
Accusations fly: ruling parties in India and Bangladesh allegedly deployed AI for propaganda; opposition in Pakistan used it creatively.
Parties embraced AI legally for outreach—Modi’s speeches translated via AI reached non-Hindi speakers. Negative uses occurred: manipulated clips defaming rivals in India, voter suppression fakes in Bangladesh/Pakistan. Platforms removed thousands of accounts; fact-checkers flagged cases.
Contextually, South Asia’s politics feature “IT cells” spreading rumors pre-AI. Generative tools scale this, but regulations lag. Ethical angle: AI levels fields for resource-poor campaigns, yet enables unchecked smears—hypocrisy in decrying opponents’ use while employing similar tactics.
Verdict: Misleading. Parties used AI extensively, some harmfully, but mostly for amplification, not unique destabilization.
Claim 4: Low Media Literacy in South Asia Made AI Misinformation Especially Dangerous, Leading to Widespread Belief in Fakes
With varying literacy and high social media use, fears arose that voters would accept deepfakes as real, eroding trust.
Vulnerable groups exist, but resilience shows. Fact-check networks in India debunked virals quickly; WhatsApp forwards often carried warnings. Exposure high (e.g., 75 percent in India), but belief lower—many recognized glitches or context. Platforms’ labels and removals helped.
Philosophically, information ecosystems evolve; colonial-era rumors spread without tech. Trade-off: AI exploits trust in shared content, but boosts literacy efforts—fact-check helplines gained traction.
Verdict: Misleading. Risks heightened vulnerability, but widespread unquestioning belief unproven; countermeasures limited damage.
Claim 5: AI Misinformation Destabilized Elections by Eroding Trust in Results and Institutions
Post-poll claims suggest fakes fueled boycotts or disputes, questioning legitimacy.
Elections faced controversies—rigging allegations in Pakistan/Bangladesh predated AI. Outcomes aligned with pre-poll expectations; no AI-linked annulments like some global cases. Trust issues stem from polarization, suppression—not primarily AI.
Wider implications: AI adds to “liar’s dividend”—politicians dismiss real evidence as fake. Yet 2024 showed restraint; malicious uses minor versus predicted “tsunami.”
Verdict: False. Contributed to noisy environment, but not primary destabilizer.
AI-generated misinformation marked South Asian elections with innovative campaigns and scattered deceptions. Tools reached linguistically diverse voters positively while enabling smears. Yet measurable destabilization—shifted outcomes, widespread violence, collapsed trust—lacked evidence. Traditional misinformation, suppression, and divides drove tensions more.
Contradictions highlight: AI democratizes speech but risks abuse; panic may overregulate beneficial uses. Ethically, platforms and parties bear responsibility—transparency in AI content aids trust. For South Asia’s young democracies, real threats lie in offline inequities. AI amplifies voices; guarding democracy demands facts over fear.
