Deepfake Videos of Indian Politicians and Celebrities Are Exploding Ahead of 2026 — Can You Even Trust What You See Anymore?

In January 2024, a former chief minister of Tamil Nadu appeared on video at his party's youth conference. Same yellow scarf. Same white shirt. Same familiar, slightly tilted stance he was known for. The only problem Muthuvel Karunanidhi had died six years earlier, in 2018. His own party had authorised the deepfake, bringing him back for one more appearance to energise the crowd.

A month later, an audio clip of Jayaram Jayalalithaa Tamil Nadu's legendary "Amma," gone since 2016 showed up on her party's official social media account, her voice apparently commenting on current politics as if she'd never left. Around the same election, ordinary voters started getting phone calls from "their local representative," discussing local issues in a warm, familiar tone. Except the person on the other end of the line wasn't a person at all. It was an AI voice, trained to sound exactly like someone they already trusted.

None of this is science fiction anymore. It already happened, and it's only getting more common heading into 2026.

📌 Key Takeaways

• ✅ Google searches for "deepfake election" jumped 340% in the run-up to 2026 across India, the US, and Brazil
• ✅ Dozens of cheap apps can now sync a fake script to someone's real face and voice in minutes — no technical skill required
• ✅ India's IT Rules 2026, in force since February 20, now require AI content to be clearly labelled, with a 3-hour takedown window for the most harmful kind
• ✅ Women, not politicians, are actually the most common deepfake target overall — often used for harassment rather than politics
• ✅ Rural and less digitally literate voters have been shown to struggle the most at telling real footage from fake
• ✅ There are real, practical ways to spot most fakes today — this isn't a lost cause yet, even if it's getting harder every year

The Problem: This Technology Got Cheap, Fast, and Convincing All at Once

A few years ago, making a convincing fake video of someone required real technical skill, expensive software, and time. That barrier has basically collapsed. Research covering India's deepfake landscape found that dozens of apps now exist that can sync a person's lip movement to a customised script within minutes, producing low-quality but often good-enough fakes with almost no technical effort at all. What once needed a small production team can now be done by one person with a laptop and an afternoon to spare.

The scale of interest in this problem reflects how quickly it's escalated. In just the 30 days leading up to the 2026 election season, Google Trends recorded a 340% surge in "deepfake election" searches across the United States, Brazil, and India a spike that reflects genuine, widespread public anxiety, not a niche tech-world concern anymore.

Real Examples: This Has Already Happened, Repeatedly

This isn't a hypothetical future risk it's already a well-documented pattern from recent Indian elections, and it's genuinely worth being specific about how it played out, without taking sides on who was more at fault. During the 2024 general election, deepfake videos of Bollywood stars Aamir Khan and Ranbir Kapoor appeared showing them criticising the sitting prime minister, ending with an opposition party's election symbol. Both actors publicly denied any involvement, and one actor's father filed a police complaint. Around the same time, a doctored video of an opposition leader made it appear he had resigned from his own party, and a fabricated clip circulated of a sitting Home Minister discussing changes to the reservation system.

This wasn't limited to any single party research tracking the election found deepfakes were used across the political spectrum, sometimes to mock opponents, sometimes for outright disinformation, and occasionally, oddly, for humour that politicians themselves seemed to enjoy a widely shared deepfake of the prime minister dancing reportedly earned his own public praise. It's a genuinely messy, two-sided story: the same technology being used as a weapon, a joke, and a genuine act of remembrance, sometimes within the same election cycle.

It's Not Just Politicians And That Matters

Here's a detail that often gets lost in the political headlines. Research monitoring deepfakes globally has consistently found that women, not politicians, are actually the most frequent targets of this technology overall, often as a tool of harassment rather than political messaging. Indian journalist Rana Ayyub, for instance, became the target of a coordinated deepfake harassment campaign after she spoke out about a government response to a child's assault case a case that had nothing to do with elections at all.

This pattern has continued into the celebrity space too. Actress Rashmika Mandanna was the subject of a viral deepfake in late 2023, created in a genuinely mundane, almost anticlimactic motive simply to grow the follower count of a fan page. The creator was later arrested. There was no grand political agenda behind it at all just an ordinary person exploiting an extraordinarily powerful new tool for a fairly small, petty reason, which is in some ways the most unsettling part of the whole story.

The Solution the Government Has Actually Put in Place

India's response to this hasn't been purely reactive. On 10 February 2026, the Ministry of Electronics and Information Technology formally notified the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, which came into force ten days later. The rules define "synthetically generated information" for the first time in Indian law, and cut the deadline for removing flagged unlawful content dramatically from 36 hours down to just 3, specifically for content that impersonates a real person in a sexually explicit, defamatory, or violence-inciting way, or that impersonates government officials.

The labelling requirement is genuinely specific too any AI-generated visual content must carry a label covering at least 10% of the visual surface, or the first 10% of the audio duration, and platforms with more than five million users must collect a declaration from every uploader about whether their content is AI-generated. Miss the takedown window, and a platform risks losing its legal safe harbour protection entirely exposure that, combined with related data protection law, can reach penalties of up to ₹250 crore per breach.

My Honest Take: The Law Is a Good Start, But It's Fighting an Uphill Battle

I genuinely think the IT Rules 2026 are a serious, well-designed first attempt at this problem — a 3-hour takedown window and mandatory labelling are exactly the kind of concrete, enforceable measures that were missing before. But I don't think anyone should assume this quietly solves the problem, and here's why.

A law that requires labelling only works on platforms willing to enforce it, and on creators who bother to comply in the first place. The apps making these deepfakes cost next to nothing and take minutes to use the barrier to creating one has essentially disappeared, while the barrier to catching and removing every single one, especially the ones shared quietly in closed WhatsApp groups rather than public platforms, remains genuinely high. I'd bet a meaningful share of the deepfakes actually circulating during the 2026 election season never touch a platform large enough to fall under the five-million-user labelling requirement at all. WhatsApp forwards, small regional Facebook pages, private groups a lot of the real damage in past elections happened exactly there, in the spaces hardest for any single law to reach.

What worries me more than the technology itself, honestly, is what research keeps finding about who actually gets fooled. Studies of the 2024 election found that voters with less than a high school education, and those in rural settings, struggled the most to tell real footage from fake. A law can label a video. It can't, on its own, teach someone who's never had a reason to doubt what they see on a screen to suddenly start questioning it. That gap — between what the law can technically require and what an average, busy, trusting viewer will actually notice — is where I think the real risk still sits.

How to Actually Spot One Yourself

Waiting for a platform to label something isn't always fast enough. A few practical checks genuinely help, even without any special tools.

Watch the mouth and the eyes closely, not the overall scene. Lip-sync mismatches, unnatural blinking patterns, or a mouth that moves slightly out of rhythm with the audio are still common giveaways in cheaper deepfakes, even convincing-looking ones. Pause the video on a still frame during a fast head turn — many current tools still struggle with motion blur and produce a slightly warped or smeared face for a split second.

Check who else is reporting it, not just where you saw it. A genuinely major political statement from a real public figure will be independently reported by multiple established news outlets within hours. If a shocking clip is circulating only through forwarded WhatsApp messages or a single social account with no other outlet carrying the same story, that absence is itself a meaningful signal.

Use a dedicated detection tool for anything you plan to act on or share widely. Indian fact-checking platform BOOM has used itisaar, a deepfake detection tool built by IIT Jodhpur, to successfully identify AI voice clones in past viral videos. Tools like this are increasingly accessible and worth reaching for before forwarding anything that seems designed to provoke an immediate, strong reaction.

Be extra cautious of content that's engineered for maximum emotional reaction. Deepfakes built to mislead are rarely calm or neutral — they're built specifically to make you angry, afraid, or eager to share instantly. That emotional urgency is often the giveaway itself, regardless of how convincing the actual footage looks.

Where This Leaves Us

The honest answer to "can you even trust what you see anymore" is a little uncomfortable — not automatically, no, not the way you could a decade ago. But that doesn't mean the situation is hopeless, or that every video needs to be treated with total paranoia. It means the old habit of "seeing is believing" needs a genuine update, for pretty much everyone, regardless of how tech-savvy they consider themselves. The law is finally catching up. The detection tools are getting better too. What hasn't caught up yet, for a lot of people, is the instinct to actually pause before believing — and forwarding — something that feels a little too perfectly designed to make you feel a certain way.

Frequently Asked Questions

How common are political deepfakes in India right now?

Genuinely widespread and growing. During India's 2024 general election, deepfakes were used across the political spectrum — including fabricated celebrity endorsements, a doctored resignation video of a major opposition leader, and even AI-generated appearances by deceased politicians authorised by their own parties. Google Trends data recorded a 340% surge in "deepfake election" searches across India, the US, and Brazil heading into the 2026 election season, reflecting how mainstream this concern has become.

What new laws exist in India to deal with deepfakes?

The IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, notified in February 2026, formally define "synthetically generated information" for the first time and require mandatory labelling of AI-generated visual and audio content. The rules also introduce a strict 3-hour takedown window for the most harmful deepfake content, such as material that is sexually explicit, defamatory, or impersonates government officials, with platforms risking loss of legal safe harbour and penalties of up to ₹250 crore for non-compliance.

Are politicians actually the most common deepfake targets?

No, research consistently shows women overall are the most frequently targeted group by deepfake technology, often for harassment rather than political purposes, rather than public officials specifically. High-profile political and celebrity cases tend to get the most media coverage because of their visibility and viral reach, but the broader pattern of deepfake misuse extends well beyond politics into personal harassment and reputation attacks against private individuals.

How can someone tell if a video is a deepfake?

Useful signs include lip-sync mismatches, unnatural blinking, or facial distortion during fast movements, since many current tools still struggle with motion. Checking whether a major claim is independently reported by multiple credible news sources, rather than circulating only through forwards or a single account, is also a strong signal. Dedicated detection tools, such as IIT Jodhpur's itisaar platform used by Indian fact-checkers, offer another reliable way to verify suspicious content before sharing it.

Who is most vulnerable to being misled by deepfakes in India?

Research on the 2024 election found that voters with lower levels of formal education and those in rural areas struggled the most to distinguish real footage from AI-generated fakes. This gap highlights that legal protections and detection tools alone aren't sufficient — genuine public awareness and digital literacy remain critical, particularly for the audiences least equipped to independently verify what they're seeing.

This erosion of trust in what appears online connects to other ways AI is reshaping everyday safety and judgment in India. The New Scam: AI Is Cloning Your Voice to Fool Your Own Parents looks at a closely related use of this same technology, and Finfluencers: Why the Financial Advice on Your Instagram Feed Might Be Wrong covers another corner of the internet where what you see isn't always what it claims to be.