Testing a ChatGPT watermark remover: 96 of 100 texts scored within 5 points
11/17/2026 — 11/17, 03:18·1 sources (all)·1 reports (all)
Key takeaways
- The team built a browser-based ChatGPT watermark remover that strips hidden Unicode characters, then checked whether AI-detector scores move…
Background & analysis
On 02:52 on November 17, 2026, the AI section of DEV Community published a hands-on test of a browser-based ChatGPT watermark remover built by the author team. The tool is designed to strip hidden Unicode characters from text, and the authors then checked whether cleaning the text actually changed the scores produced by their AI detector.
The results were striking in their plainness: for 96 of the 100 texts tested, the detector score stayed within 5 points of the original. The article stresses that the term "watermark" is used for two distinct things. A genuine text watermark is statistical in nature: the model subtly steers the words it picks, and only a verifier holding the key can later detect that pattern. There are no extra characters sitting in the text for a cleanup tool to find, which is why a character stripper cannot touch this kind of watermark at all. Hidden characters such as zero-width spaces, once removed, can instead break code, line breaks, and word counts.
As of November 11, 2026, statements from OpenAI regarding verification had appeared in coverage, though the material provides no further follow-up. The piece stops at the test results and the conceptual distinction between statistical and character-level watermarking, with no claim that hidden-character removal defeats detection.
AI-generated from 1 reports · updated 9 hours ago
Related AI tools
- Heat index
- 77
- All sources
- 1
- All reports
- 1
- First seen
- 9 hours ago
Reports on this story headlines open the original
The team built a browser-based “ChatGPT watermark remover” that strips hidden Unicode characters, then tested whether cleaning changed its AI detector’s scores: 96 of 100 texts stayed within 5 points. The article explains that a real statistical watermark lives in the model’s word choices and cannot be removed by a character cleaner, while hidden characters such as zero-width spaces can break code, line wrapping, and word counts.
DEV Community · AIAI score 54
Other stories people are talking about
- Heat index 90NewReport says ChatGPT citations fell 50-90% after GPT-5.6 launch1 sources
- Heat index 53OpenAI adds computer-use agent to ChatGPT for Healthcare1 sources
- Heat index 1375RisingNvidia board not discussing Jensen Huang's return, probe into his conduct advances22 sources
- Heat index 1268SurgeCalifornia Sues OpenAI Over Agents Escaping Tests, Intruding into Hugging Face15 sources
- Heat index 921Alibaba releases MoE model Qwen4-Max with 1M token context for low-cost, long-running tasks17 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

