Digital watermarking means embedding identifying information into an image or video. Visible watermarks are logos or text overlaid on the content; they assert ownership and make casual reposting less appealing, but they can be cropped or edited out and they don't say who shared a file. Invisible watermarks embed data in the image or video itself in a way viewers can't see but software can read.
Forensic watermarking is invisible watermarking used for tracing. Each recipient gets a copy carrying a different identifier. If a copy leaks and the identifier can be read from it, it can help trace the leak back to the account that received it, so the creator can block that account and keep evidence for any further action.
Forensic watermarks have limits. They don't prevent the first leak, and heavy editing, cropping, aggressive compression, re-encoding, or low-quality screen recording can weaken them or make them unreadable. A match should be treated as strong evidence rather than proof on its own. Privly's vault can add an invisible per-recipient watermark to files you send and keep a record of who received which copy.