# Can I watermark my own writing so I can prove it is mine later? Source: https://global100.org/forum/q/can-i-watermark-my-own-writing-so-i-can-prove-it-is-mine-later/ Site: Global 100 Forum, category Provenance and C2PA Published: 2026-11-04. Updated: 2026-11-05. Replies: 7. ## Question by Lena B., 2026-11-04 After the false flag a few weeks ago, I want a better answer than "trust me" the next time a client questions a piece. I keep reading about watermarks and content credentials, and they sound like a way to stamp my work as mine. So here is my question in the simplest form. Can I add some kind of watermark or signature to plain text that I write, so that later I can show it existed and came from me? I mostly deliver documents and web copy. Not images, not audio. What I have in mind, in order of how technical it gets: an invisible mark in the text, a signed or timestamped file, or an official registration. I have seen people mention all three and I cannot tell which are real and which are wishful. I also want to know what would hold up if it came to a dispute. A client saying "your text scored as AI" is one problem. A client saying "you took this from somewhere" is a different problem. Does one tool cover both? I do not need courtroom grade. I need something a reasonable person would accept, and I would like not to spend a week on it. ## Reply 1 by Jules, 2026-11-04 you can hide a watermark in the text with invisible characters, zero width spaces and that kind of thing. put a pattern of them in your documents and it is a signature only you know. then if the text shows up somewhere you can check for your pattern. simple and free. ## Reply 2 by Hana Sato, 2026-11-05 @Jules zero width characters are removed or altered by many editors, forms and publishing systems, and they disappear if someone retypes or reformats the text. Anyone who strips formatting also strips them. They can also make a document behave oddly, for example in search, spell check or a code editor. I would not call that a signature. It is a mark that survives only until the first careless copy and paste, and it proves nothing about who put it there. ## Reply 3 by Marco Ferri, 2026-11-05 Content credentials, the C2PA approach we use for photos, are built around media files such as images and video. They do not give you a signature for plain prose in a document, which is what you deliver. The [content credentials thread](/forum/q/what-is-c2pa-and-do-content-credentials-actually-help/) is worth reading to see what that system is for and what it does not cover, but it is not your tool. What carries over from my world is the habit: keep the original, with its dates, and be able to show the chain from first draft to final. ## Reply 4 by Kwame A., 2026-11-05 Keep two questions apart. Showing who wrote a text and showing who owns it are different, and different tools help with each. On ownership, in the US copyright arises automatically when the work is fixed, and registration is optional but has real advantages. On "I wrote this myself rather than with a chatbot", copyright says nothing at all. Registration records a claim to a work. It does not prove how you made it. ## Reply 5 by Sam Whitlock, 2026-11-05 Practical and cheap: put the final file in a version-controlled folder, or hash it and store the hash somewhere with a trusted timestamp, such as an email to yourself or a public commit. That shows a particular file existed in a particular form at a particular time. It does not show who typed it, and it will not impress a sceptical lawyer by itself. But combined with dated drafts it makes a coherent story. Think of it as evidence of existence, not of authorship. ## Reply 6 by Tomas Reyes (staff), 2026-11-05 Not in the way you mean. There is no robust invisible watermark you can add to plain text that proves authorship, and the watermarks in the research literature work on the machine side. Your best evidence is the paper trail: dated drafts, notes and versions. For ownership, copyright already applies, and registration is available if you want it. The research watermark most people have heard of is described in ["A Watermark for Large Language Models"](https://arxiv.org/abs/2301.10226). It works by choosing a randomised "green" list of tokens before each word is generated and softly promoting those tokens during sampling, so that a statistical test can later detect the pattern from a short span of text. That is a mark placed by a model while it generates, and the authors present it for proprietary language models. It does not apply to a human typing in a document, so it cannot sign your work. On the legal side, the U.S. Copyright Office's ["Copyright in General" FAQ](https://www.copyright.gov/help/faq/faq-general.html) says that a work is protected the moment it is created and fixed in a tangible form, registration is generally voluntary, but you must register to bring a suit for infringement of a U.S. work. Registration within five years of publication is treated as prima facie evidence in court, and registered works may qualify for statutory damages and attorney's fees. The same FAQ says that mailing a copy of your own work to yourself, sometimes called a "poor man's copyright", has no basis in copyright law and is not a substitute for registration. So the plan is: keep a dated record of how each piece came into being, store the final files somewhere with timestamps, and register only the work that is worth the effort. None of that proves you did not use a chatbot, which is a question for the process evidence and not for a stamp. Jurisdictions outside the US differ, so check yours. ## Reply 7 by Lena B., 2026-11-05 @Tomas Reyes that is the clear answer I needed. Dated drafts first, timestamped finals second, registration only for work that matters. And no, I will not be adding invisible characters to anything. Thank you. --- Cite as: Global 100 Forum, "Can I watermark my own writing so I can prove it is mine later?", https://global100.org/forum/q/can-i-watermark-my-own-writing-so-i-can-prove-it-is-mine-later/, accessed 2026-11-10.