Rahul, 35, a senior consultant on Janata Dal (United)’s social media team, spends much of his day checking Facebook and Instagram insights. The numbers tell him how voters are responding to the party’s election campaign videos: Chacha Chaudhary and his trusty sidekick Sabu drifting past flyovers and freshly laid roads in Patna.
In these clips made using Artificial Intelligence (AI), the comic book heroes are celebrating Nitish Kumar’s governance.
“They are getting a good response,” Rahul said, proud of the technology powering his party’s digital campaign. In one video, Chacha Chaudhary tells Sabu: “Bihar will no longer be known for potholes, but for new heights,” as they zip past a new flyover. Hashtagged #DhanyawadNitishKumar and #NitishKaVikasModel, the clip has drawn over 4,00,000 views. Nowhere on it is a watermark of ‘AI-generated’.
The idea, Rahul explained, was born during the team’s daily brainstorming sessions. “We created around 16–17 videos in the Chacha Chaudhary series, each showcasing a different aspect of Bihar’s development,” he said. “The way AI has evolved, we can now represent what we imagine in a graphic format,” Rahul said.
The AI Chacha Chaudhary was followed by another AI series: Changu and Mangu, caricatured versions of Rahul Gandhi and Tejashwi Yadav, who tour Bihar only to marvel at Nitish’s “development works and law-and-order situation.”
None of these was outsourced. Instead, the in-house team, armed with an internet connection and a handful of free tools, comprising 7-8 video editors and a 100-member social media team, produced it themselves.
The decision to build capacity internally, Rahul added, was strategic: “People hired from the market usually stay for the short term, but our members have been with us for years. That has built trust.” The team received some initial training from consultants, but now, most of the execution is done internally. Rahul estimates that their memes and animated ads now reach nearly 70 million unique users a month.
JDU's AI-generated Chacha Chaudhary and Sabu.
Bihar has never lacked spectacle. But this time, comics to caricatures, WhatsApp forwards to videos, political parties have the same realisation: Bihar 2025 is the state’s first AI-driven election.
Beneath the novelty lies a sharper dilemma—politicians appear to speak in dialects they never recorded, comics are reborn as propagandists, and memes blur the line between humour and propaganda. Voters now see and hear all of this without knowing whether it’s real or machine-generated.
Old Tactics, With AI
Anmol Sovit, the social media head of the Bharatiya Janata Party (BJP)'s Bihar unit, keeps his phone close; its screen flashes with hundreds of messages across dozens of WhatsApp groups. Each group corresponded to a single polling booth. “We have WhatsApp groups at each booth level, which helps us send targeted messages to specific booths,” he said.
For the party, AI is less about flashy characters and more about scale. “BJP is training its workers to create AI-generated videos,” Sovit said. The training uses only free, publicly available tools.
These training sessions are embedded in their regular team meetings. Volunteers learned to use ChatGPT to design image cards and short videos. “The purpose is to ensure our volunteers are not left behind. Even a basic understanding of technology can help them, not just in organisational work but also personally,” said Sovit. The workshops even included reminders about the IT Act, so workers “don’t misuse AI and land themselves in trouble”.
Somesh Pandey, co-convener of the Bihar BJP IT Cell, said the focus has been on practical demonstrations. “We showed workers how to generate reels with simple prompts—for instance, transforming Lalu Prasad Yadav into an eight-year-old for satire, and then adding voice effects,” he explained. The idea, he added, was to keep content fresh: “The public gets bored if they see the same kind of posts again and again. With AI, we’re trying new things, and the response has been good.”
Initially, some workers found the process intimidating. But once broken down step by step, Pandey said, “they started enjoying it—especially those already active on social media.”
The Rashtriya Janata Dal (RJD) insists on being the “pioneer” of AI campaigns. “We were the first party in Bihar to start using AI in our messaging, in songs and videos,” said a member of its social media team. Their model is modular: short, seven-to-eight second AI clips stitched together into reels. Tools include Grok for graphics and Google’s VEO for videos.
“Back in 2014 or even 2019, a photo with a caption was enough. Now it’s the era of AI,” he added.
Congress, meanwhile, has gone for satire. “This time we introduced a monkey character in our campaign, and it’s getting a good response,” said Saurabh Sinha, the party’s social media head in Bihar. “The monkey represents the young generation—dressed in a T-shirt and pants, targeting first-time voters.”
The monkey mascot reels have taken off on Facebook, each crossing at least a million views. One in particular—on ‘Vote Chori'—has gone viral, racking up nearly 4 million views and 20,000 shares. The comments range from laughter at the quirky Bhojpuri song to sharp digs at the Election Commission. In the reel, the monkey sings a line that loosely translates to “ECI is a vote thief,” alongside caricatures of PM Modi and Election Commissioner Gyanesh Kumar shown in cahoots.
Cheap, Fast, Viral
Congress' AI monkey mascot.
Beyond novelty, the real driver is cost. Consultants agree the appeal is efficiency at scale. Raj Abhishek, director of the consultancy Political Hub, has seen the economics firsthand. “WhatsApp messaging that used to cost 80 paise per message is now just 6 paise,” he said.His company is making AI-generated videos of Prime Minister Narendra Modi addressing voters in local dialects. These are sent directly to people’s WhatsApp inboxes. “We also use AI voice cloning in regional languages, with scripts developed after several back-and-forth discussions between the party and our team,” he said.In Jharkhand, he recalled, this approach was especially impactful. “It created an emotional bond,” he said. The BJP gained in those areas, the consultant noted.“We can now make a one-minute video for just 2,000 to 3,000 rupees. Earlier, we had to find local actors, pay them, shoot the video, handle post-production, and create the music. That entire process used to cost three times more,” the RJD worker also said.
Congress, too, has leaned on AI to sharpen its attacks. For instance, it used synthetic graphics to highlight the closure of a sugar mill in Motihari and targeted the government over the CAG’s report on a Rs 70,000-crore scam. “Our view is that the more new technology we use, the more it will grab people’s attention,” said Sinha, acknowledging Congress was a late entrant compared to RJD.
Consultants and the AI Playbook
While IT cells flood timelines with memes and reels, consultants are running the machinery of Bihar’s AI campaign. Firms like Rajyatantra, Leadtech, and Political Hub are using AI not only to generate content but also to measure sentiment, predict turnout and micro-target voters.
Rajyatantra has launched a dedicated campaign platform called Arthashastra, pitched as an all-in-one dashboard. It integrates social listening, API-driven data scraping and booth-level analysis. “We use historical booth-level data,past performance, turnout, and sentiment patterns and compare these with human assessments,” explained Madhan, who heads the platform. “What once took days in Excel now takes under an hour. But the final call is always made by humans.”
Leadtech, by contrast, focuses on voter surveys through an AI-powered Interactive voice response (IVR) system. Pre-recorded synthetic voices dial voters with scripted questions, record replies and convert them into text. But the limitations are obvious.
“AI voices cannot detect nuance—for example, if a respondent says they will vote BJP for its ideology but are unhappy with performance—but AI still has no context,” said Vivek Singh Baghri of Leadtech.
Thus, humans still have to sift through calls.
Sampling, too, is fraught without human intervention. “If a constituency has 200 Yadavs, 300 Brahmins and 200 Thakurs, the sample must reflect these proportions. Without that, survey results are misleading,” said Baghri. The response rate is another challenge: “Fifty to sixty percent of people either don’t answer or hang up once they realise it’s not a human caller,” he added.
The data that fuels the drive come from old-fashioned ground work. Campaign workers distribute calendars or pamphlets with a leader’s photo and collect mobile numbers in exchange which are then added to databases for follow-up calls and WhatsApp blasts. Consent, as Baghri admitted, is loosely defined.
Most of the AI services powering this backend are outsourced to third-party providers rather than developed in-house. For analytics, consultants use tools like Sprout Social and Hootsuite Insights; for videos, Pixverse and Runway AI; for graphics, Gemini and ChatGPT; and for voices, ElevenLabs and India TTS.
This patchwork of outsourced tools makes campaigns agile, but it also dilutes accountability. These services were built for marketing or entertainment, not politics—and none were designed with electoral transparency in mind.
Between Innovation and Manipulation
The lack of transparency is a thread running through every stage of Bihar’s AI experiment. Raj Abhishek admitted that their AI-generated videos carry no watermark, making it impossible for an ordinary voter to distinguish between synthetic and authentic content.
“That is part of the risk—voters consuming such material often cannot distinguish between real speeches and AI reproductions,” he said.
To address this, Political Hub has hired fact-checkers on contract. Their role is to verify information before circulation and to counter fake messages already spreading in a particular region. But even Abhishek conceded that detection remains difficult: AI voices and visuals have become so convincing that it is difficult to separate fact from fabrication.
Other firms have tried to self-regulate. Rajyatantra says it maintains audit trails of AI-generated creatives and stores them in controlled folders. “We disclose their source when used on social media,” the company said, while adding that “future regulations should mandate clear labelling of AI-generated content.”
But at present, disclosure is voluntary. There are no binding rules that force parties to label synthetic content—and that leaves voters navigating a flood of AI-driven narratives with little clarity over what is real and what is fabricated.
The Regulation Gap
Unlike countries where laws on AI in elections are already emerging, India is stuck in a policy vacuum. In the U.S., states have passed laws requiring disclosure of AI in political ads or banning deceptive deepfakes.
In India, the Election Commission’s steps are limited to advisories. The ECI has urged parties to label AI-generated content with tags such as “AI-Generated” or “Synthetic Content,” particularly in materials submitted for certification during the Model Code of- Conduct (MCC) period. In May 2024, it also directed parties to take down deepfakes within three hours of notice. But enforcement is patchy: outside the MCC window, these rules carry no real weight.
Social media platforms, too, have no binding framework. “Facebook and YouTube are now promoting less AI content, so for now we have reduced its use,” Congress’s Saurabh Sinha acknowledged. But these are just algorithm tweaks, not enforceable standards—and whatever moderation exists differs from platform to platform.
The result is a regulatory gap that parties and consultants are exploiting. They can make AI content with easy tools, strip away watermarks, and circulate synthetic voices or images long before the MCC applies.
