How do you check if your likeness is being used in AI content?

As of 24 October 2026: the service descriptions, prices and turnaround figures on this page are current.

What are the four checks, and what does each one catch?

Four, in ascending order of power. Reverse image search on a single frame of the suspicious media finds direct copies, and it is fast and free, but it misses a re-cropped, re-encoded or re-generated version. Perceptual hashing is the same idea done properly: a hash of the visual content, not the file, so that a resized, watermarked or screen-recorded copy still matches, and it is what finds the same synthetic video re-uploaded under three different names. A forensic comparison against a reference pack of your own dated images measures the match rather than guessing at it, and produces a written finding you can attach to a complaint. A naming and media sweep covers the case where the content carries your identity but not your face: your name on a synthetic profile, a cloned voice in a video, an endorsement you never gave.

What should be in your reference pack?

Twenty to thirty images and clips, captured in one sitting and dated, covering the conditions an attacker would struggle to reproduce: front-facing, three-quarter and profile angles; different lighting, including flat indoor and strong side light; neutral expression, speaking, smiling and mid-blink frames; and a range of ages and appearances, since a deepfake trained on a decade-old photo is a different problem from one trained on last week's reel. Store them somewhere immutable with the capture date and a hash of each file. Without a reference, a forensic comparison has to argue from general appearance, which is a weak position in a complaint. With it, the analysis can state a match confidence and point to the specific frames where the synthetic image departs from the real one.

What should you do the moment you find a match?

Four things, in this order, and the order is what makes the takedown work. Capture the evidence first: the URL, a full-page screenshot showing the address bar and a visible timestamp, the downloaded original file with its hash computed, the account details and the surrounding context. Then file on the platform the same hour through the IT Rules 2021 grievance route, attaching the media file rather than a link, because the February 2026 amendments set a 3-hour window for flagged unlawful content and a platform will not act on a broken link. Then report on cybercrime.gov.in and helpline 1930 if the content is impersonating you, is non-consensual, or is fraudulent. And then watch for the re-upload, because a removal without a watch is a two-week reprieve at best, and a monitored watch finds the next copy in hours.

How often should you check, and what does it cost?

A monthly routine, if you do nothing else. Two hours, in four parts: a reverse image search on your own name and on three reference frames, a sweep of the platforms you actually use for profiles carrying your display name, a search of video platforms for your name, and a check of whether any of it has been reported already. A one-off professional sweep, which is what most people want immediately after an incident, is Rs 15,000 to Rs 40,000 excluding 18% GST and covers the reference pack, the forensic comparison, the platform search and the first takedown filings. A monitored watch, which is what a company or a public figure should hold, is Rs 12,000 to Rs 30,000 per month and adds continuous detection plus automatic takedown. For a person who has never had a problem, the monthly routine is proportionate; for a company that has had one, the watch is the only thing that holds.

What are the frequently asked questions?