Frequently asked questions
- What is ZeroTrace AI?
- A text hygiene tool. It finds and removes invisible Unicode characters, steganographic payloads and hidden-text prompt injections from any text, and reports every change it made. It is available as a web playground and a REST API.
- Does it remove AI watermarks?
- No, and no character-cleaning tool can. Google's SynthID-Text and Anthropic's watermark for Claude are statistical: they live in which words the model chose, not in any character added to the text. Anthropic's documentation states plainly that nothing is added and there are no hidden characters. Removing invisible characters has no effect on them.
- Is there a ChatGPT watermark I can remove?
- No. OpenAI built a text watermarking system but never deployed it. The narrow no-break space (U+202F) that circulated as a supposed watermark is a formatting artifact that encodes nothing.
- Will it help my text pass an AI detector?
- Generally no. AI-detection services classify writing style and sentence structure; they do not examine invisible characters. Cleaning typography can reduce scores on naive heuristics that key on smart quotes and em dashes, but that is not the same as defeating detection, and we will not claim otherwise.
- What can it actually detect?
- Zero-width and invisible characters, hidden text smuggled through the Unicode Tags block, StegCloak-style and base-N zero-width steganography, variation-selector payloads, bidirectional text attacks such as Trojan Source, whitespace steganography in trailing spaces and tabs, mixed-script homoglyphs, and typographic artifacts such as smart quotes and em dashes.
- Will it break text in other languages?
- No. The default safe profile keeps zero-width joiners inside emoji sequences and between letters of Arabic, Persian and Indic scripts, keeps variation selectors on CJK ideographs, and keeps the tag sequences that form the England, Scotland and Wales flag emoji. Turkish, Greek, Cyrillic and accented Latin letters are ordinary characters and are never touched.
- Is my text stored?
- No. The text you send is never written to a database or a log. Only counters are recorded: character count, how many findings there were, which categories they fell into, and what the request cost.
- How much does it cost?
- The playground is free without an account, up to 5,000 characters per request. A free account gets 300 credits a month and 30,000 characters per request. Pro is $9 a month for 3,000 credits, Scale is $29 a month for 15,000. Credit packs start at $5 for 50 credits. One credit covers up to 10,000 characters.
- Is there an API?
- Yes. POST to https://api.zerotraceai.net/v1/clean with an API key in the Authorization header. There is also /v1/detect, which returns the findings without the cleaned text. API keys are unlimited per account and rate limits apply per account rather than per key.
- How many characters does it check for?
- 143,868 code points: the full Unicode 17 set of Default_Ignorable, format, control, whitespace, bidirectional, private-use, surrogate and noncharacter code points. The four left out are tab, line feed, carriage return and the ordinary space, where the signal is the pattern rather than the character — handled by a separate whitespace steganography detector.
More detail in the guides, and the full list of what this tool cannot do on the honesty page.