AI is, currently, right now, at this very moment, invading all of human writing. Each passing month, each new model, each increase in processing power, the end result seems more and more certain — AI intelligence is increasing, AI skillsets are growing, and the gap between AI and human output is shrinking. Already, AI can write technical copy at qualities equal-to or better-than the majority of human beings. Large Language Models are practically purpose built for this one skillset — communicating information through written language.
At the same time, creative writing platforms like Royal Road and Wattpad are in the midst of an AI content crisis. Hundreds, potentially thousands, of AI written projects have flooded the platforms, muddying site content and blurring the lines between human creativity and… yes, I'll say it — AI slop.
This problem will only continue to amplify as models become more and more powerful and efficient.
It's not the only medium being overrun. We saw it three years ago with image generation, we're seeing it right now with computer programming and video generation, and its hard to accurately picture the state of the world we'll be living in five years from now.
By 2030, will we have ceded all human content generation to AI? Will the world be flooded with such an endless wash of AI content that human output is dwarfed by comparison, drowned on all fronts?
WriterMark is one tool in a ever-more-necessary toolbox for protecting the creativity and originality of human kind. WriterMark is a system of telemetric monitoring designed to explicitly verify a body of text as "Human Verified" — meaning it was likely written by a human. Much like Google's Recaptcha, it will not be a perfect, be-all-end-all solution, but I do believe that telemetric verification will play a crucial role in assessing the originality of a written piece of work.
WriterMark is a system of verification for writing that watches the process of writing, rather than checking the content of the text. This system is fundamentally different from other AI checking tools like Grammarly or — (tools which are already failing to identify AI generated text). That's because AI generated text is not formally different from human text — it uses the same words, the same construction of sentences; it can be guided, prompted, adjusted, made to seem more human. Whats much harder to falsify is the act of writing the text on a keyboard.
WriterMark collects anonymous telemetric data from your writing session as you type, testing that data against an algorithm designed to recognize key indicators of human typing. WriterMark runs in continuous cycles, monitoring the entire length of a writing session and associating the score with the relevant document. Every time the WriterMark system triggers, it does the following things.
It means that despite the WriterMark server never seeing what you wrote, and having no access to your actual text content, you can later prove that the exact text content of a written piece is associated with the exact "Human Verified" certificate you received. No guessing, no "trust-me's", no faking a certificate. At any time, you can take your text, take your locally provided token, put them both into the WriterMark verification system, and prove that the score you received was for the exact body of text you wrote. Need to edit your text? No problem. The system runs again, re-certifies your content using the certification system, and you once again have a certificate of proof.
WriterMark additionally confirms that the session timeline seems realistic, meaning you can't have an AI simply batch out an array of fake telemetry and have it pass. WriterMark confirms the session took the time it claims to have taken, and compares this against realistic human capabilities for typing speed.
This is a near-instant solution for the lowest level of AI fakery — having ChatGPT write you some text content, and pasting it directly into a text editor. WriterMark will fail any text content that is majority copy and paste from an external source.
Copying from external sources cannot be verified, meaning anything generated from an AI or written without verification will fail. Copying within a WriterMark document or between WriterMark documents will work seamlessly, however. The WriterMark system builds a map of generic edit information that matches one-to-one with the origin of text within the document. If you copy and paste within a human written document internally (aside from edge cases like repeatedly copying single or short strings of characters), WriterMark will not lower your verification score. The text is treated as human written, internal text, as expected. If you copy between documents, WriterMark carries the map partial from one document to another, meaning the entire system is one self contained system of verification. This prevents any and all types of copy and paste injection vectors.
With all of this functionality working together, the result is the single most powerful digital tool in existence today for proving that text was written by a human. Right now, writing a document with WriterMark verification enabled genuinely proves, beyond nearly any doubt, that the text is human written. As AI agents and AI tools become more advanced, WriterMark will need to evolve to match more difficult verification challenges (real-time AI writers, complex human mimicking bots), but that is true no matter the paradigm. What's most important is that we begin the process of protecting and verifying human writing as early as possible, and build humanity the most effective possible tools for proving that text is human written. I believe WriterMark has an important part to play in that process.
WriterMark is available in two separate apps currently:
Currently, WriterMark is enabled by default in it's sister application — Wintertext (available at www.wintertext.com right now, completely free). Wintertext is a free text editor built specifically for writing human verified documents. Wintertext shows your certification in real time, displays your certificate, lets you copy and paste your certificate anywhere, and provides a helpful link for directly verifying your text manually at www.writermark.org/verify. Wintertext documents are written and stored locally in an open file format (.wtxt), a simple JSON wrapper that packages up your text with the certificate directly embedded, along with styling information and metadata. You can also export to Markdown, JSON, HTML, and Word (.docx).
Sondernote is an alpha-stage fiction and long-text publishing platform for the web designed for publishing written content in the modern world. Sondernote offers a sleek, kindle-friendly reading experience with dynamic pagination, user accounts, comments, likes, content feeds, genres, multi-chapter works, and more. With the massive influx of AI content flooding platforms like Royal Road and Wattpad, Sondernote additionally offers optional Human Verification filters for showing only human-written content. Sondernote additionally has a native editor that, similar to Wintertext, provides real-time human verification.
End session.