How to Spot AI Deepfake Scams in India (2026 Safety Guide)

A few years ago, a convincing deepfake needed a research lab. Now it needs a phone, a few minutes and a handful of photos pulled off a public social media account. That shift is why a deepfake scam in India no longer looks like a technology story. It looks like a video of a familiar business figure recommending a trading app, a short video call from a number claiming to be your nephew, or a morphed photograph attached to a demand for money.

This guide covers the patterns actually circulating, the tells that still work, why detection tools will not save you, and the one verification habit that defeats almost every version of this fraud. It also covers reporting, including the sensitive case where morphed images are used for extortion. We cover online safety regularly at TechLein, and this is the practical version rather than the alarmed one.

Key takeaways

• Assume any video or image can be faked, and verify the person through a separate channel you already trust.
• Celebrity investment endorsements in video form are among the most common Indian deepfake scams.
• Automated deepfake detectors are unreliable on compressed social media video. Do not depend on them.
• If morphed images are used to extort you, do not pay. Preserve evidence and report immediately.

The patterns actually being used in India

Four shapes cover the overwhelming majority of cases.

Fake endorsement videos. A well-known industrialist, film star, cricketer or television news anchor appears to recommend a trading platform, a stock tip group or a quick-return scheme. These circulate as short clips on social feeds and messaging forwards, often with the audio dubbed into Hindi or a regional language. The video is usually real footage with the mouth and voice replaced.

Impersonation video calls. A short, poor-quality video call appears to show a relative in distress, or a person in uniform claiming to be from a law enforcement or investigation agency. The video is deliberately brief and low resolution, because both hide artefacts. This overlaps heavily with the digital arrest scam, where the fake official keeps the victim on a video call for hours while demanding transfers.

Morphed images used for extortion. Photographs taken from public profiles are altered into compromising images, or a brief video call is recorded and manipulated. The victim is then threatened with circulation to their contacts unless they pay.

Fake authority and offer content. Synthetic clips of officials announcing schemes, fake job offers, and fabricated news segments used to build credibility before a more ordinary financial fraud.

Scam patterns at a glance

Visual tells that still work

Detection by eye is getting harder, but artefacts remain, especially in the cheap fakes that make up most scams.

Watch the mouth against the audio, particularly on Indian-language dubs, where lip movement often matches an English original rather than the words being spoken. Look at the boundary of the face: hairline, ears, jaw and the edge of spectacle frames are where blending fails. Jewellery is unusually revealing, since a nose ring, bindi or earring will flicker or partially vanish as the head turns.

Other reliable signals: teeth that shift shape between frames, a face that stays oddly still while the body moves, lighting on the face that does not match the room, and blinking that is either absent or too regular. In generated stills, check hands, fingers, and any text on signboards or documents.

Audio has its own tells. Flat emotional range, missing breath sounds, background noise that cuts abruptly, and unnatural pacing between sentences. These matter most on calls, which our guide to AI voice cloning fraud covers in depth.

Contextual tells matter more than visual ones

Here is the uncomfortable truth: the quality of fakes improves every few months, and visual inspection is a shrinking advantage. Context does not degrade the same way.

Ask what is being requested rather than whether the face looks right. Real relatives do not insist you tell no one. Real officials do not conduct proceedings over a video call or demand payment to a personal account. Real investment opportunities do not arrive as a forwarded video with a link to a website you have never heard of. Genuine people will happily switch to a phone call or meet in person; a scammer will always have a reason why they cannot.

Any message combining urgency, secrecy and an irreversible payment is fraudulent regardless of how convincing the face on screen looks. That test has not changed in twenty years and it still catches almost everything.

Why detection tools will not save you

Several websites and apps offer to tell you whether a video is AI-generated. Their performance in ordinary conditions is much weaker than the marketing suggests, for structural reasons.

Detectors are trained on particular generation methods and degrade against newer ones. Social media platforms re-compress video heavily, destroying exactly the fine artefacts detectors rely on. And the results cut both ways: real videos get flagged as fake, which is arguably the more dangerous error, because it lets people dismiss genuine evidence.

Treat any detector output as weak circumstantial evidence. Verification through a channel you control is worth more than any percentage a tool shows you.

The verification rule that actually works

Whenever a video or call asks you for money, credentials or urgent action, verify through a second channel that you initiate.

Do not call back the number that contacted you. Do not use a link or contact detail supplied in the message. Open your own contact list and call the person on the number you have always had for them. If it claims to be an institution, look up the office independently and contact it. If it is a public figure endorsing something, check their own verified account or the company’s official site.

This single habit defeats deepfake video, cloned voice, spoofed caller ID and ordinary impersonation at once, because it does not depend on judging authenticity at all. It just moves the conversation somewhere the scammer cannot follow.

If morphed images are used to extort you

This is frightening and people often react by paying. Paying is the one response that reliably makes things worse, because it confirms the threat works and further demands follow.

Do not pay, and do not engage in negotiation. Take screenshots of the profile, messages, numbers and any payment details demanded, since these are what investigators need. Do not delete the conversation. Block the account after you have preserved the evidence, and report the profile to the platform, which is required to act on such reports.

Report it. The National Cyber Crime Reporting Portal has a dedicated category for offences of this kind, and complaints relating to obscene material involving women can be filed there without disclosing your identity publicly. You can also call the national cyber crime helpline on 1930. If you are a student, tell a parent, warden or teacher; the isolation is what the extortion depends on.

Nothing about this is your fault, and the images are fabricated. That is a fact investigators deal with routinely.

How to respond and report, step by step

1. Stop the interaction. End the call or close the chat. Nothing needs to be decided in the next ten minutes.
2. Do not transfer money, share an OTP, or install any app or remote-access software they suggest.
3. Take screenshots and note the number, profile link, timestamps and any account or UPI ID mentioned.
4. Verify independently by contacting the real person or institution on a number you already had.
5. If money has already been sent, call 1930 immediately and inform your bank. Speed matters most in the first hour.
6. File a complaint on the National Cyber Crime Reporting Portal with your evidence attached.
7. Report a fraudulent call or message through Chakshu on the Sanchar Saathi platform, which handles suspected fraud communication.
8. Report the content to the platform it appeared on, and warn the person being impersonated if you can reach them.

Frequently asked questions

Can I tell a deepfake just by looking carefully?

Often, with cheap fakes. Increasingly not with good ones. Rely on context and second-channel verification rather than your eyes alone.

Are deepfake detection apps worth using?

Only as weak supporting evidence. They fail on compressed video and produce both false positives and false negatives.

Is making a deepfake illegal in India?

Fraud, impersonation and obscene content offences apply regardless of the technology used, and platforms have takedown obligations. Report it and let the authorities characterise the offence.

What if the scammer used my own photos from Instagram?

Tighten your profile privacy and limit who can see and download your photos. Our guide to how to protect your privacy online covers the settings that matter.

Do fakes work as well in Indian languages?

Increasingly yes, though lip sync on regional-language dubs is still a common giveaway. The AI tools for Indian languages guide explains why quality varies by language.

The bottom line

Stop treating video as proof. That is the whole adjustment. A face and a voice on screen used to be reasonable evidence of identity, and in 2026 they are not. Everything else follows from accepting that: verify through a channel you opened, distrust urgency, and never let a request for money survive a hang-up and a callback.

Share the second-channel rule with the people in your family least likely to read a guide like this, because they are the ones being targeted. Our broader online safety guide for India and the guide to UPI payment fraud protection cover the financial side of the same problem, and the free AI photo editing tools roundup shows just how little effort image manipulation now takes.