About EraseGPT
Writing analysis with human judgment at the centre
EraseGPT is a Zyur product that combines AI-pattern detection, plagiarism checking and text rewriting in one web application. We build tools to support careful review and revision—not to turn uncertain signals into automatic verdicts.
Our position: an AI percentage is probabilistic, a source match needs context, and a generated rewrite needs human editing. The person using the result remains responsible for the decision and the finished text.
Who we are
EraseGPT is developed and operated as a product of Zyur. The product is designed for people who need to inspect or improve writing, including students, educators, independent writers, editors and content teams. We do not present fictional experts, invented researchers or unnamed credentials as evidence of authority.
Our aim is practical: make complex writing-analysis output understandable enough that a user can ask better questions. That means explaining what a percentage represents, separating AI detection from plagiarism, and reminding users to verify any rewritten claim before it is submitted or published.
What the product does
The AI detector returns a percentage estimate of AI-like patterns in submitted text. It does not observe how the text was written, prove authorship or identify a specific model with certainty. The plagiarism checker reports possible language overlap for source-by-source review; matching language is not by itself proof of plagiarism. The humanizer produces an editable alternative intended to improve clarity, tone and flow.
These tools can shorten the path to a useful review, but each has limits. No detector is error-free. No similarity system can search every public, private or newly created source. No rewrite can understand every fact, quotation or specialist term as well as a responsible author and subject-matter reviewer.
Our content standards
Answer first
Guidance should give a direct answer before adding detail. We use plain language and distinguish a tool result from a proven fact.
Support material claims
Methodology and educational pages should link to primary or peer-reviewed sources where those sources can clarify a technical claim.
State uncertainty
We avoid unsupported accuracy figures, guaranteed detector bypass claims and language suggesting that a score can prove misconduct.
Correct meaningful errors
We review product and educational copy as the service changes. If a reader identifies a material error, we want enough detail to investigate and correct it.
Responsible-use principles
EraseGPT should not be used to make an automatic accusation or punishment. In education, employment and publishing, reviewers should consider drafts, version history, citations, the stated rules and the writer's explanation. A meaningful appeal or correction route is especially important when an automated result could affect someone.
The humanizer is an editing aid, not a promise of “undetectable” output. Rewriting prohibited AI-generated work does not make the underlying use acceptable. Users should follow applicable school, employer, client and platform rules, disclose assistance when required, respect copyright and avoid submitting confidential material without permission.
We also encourage users to preserve the original draft, verify factual statements against reliable sources, and check that quotations and citations remain accurate after revision. Better wording is useful only when the final work is honest and sound.
Contact and corrections
For product support, privacy questions, accessibility feedback or a possible error in our published guidance, email [email protected]. Please include the relevant page URL and enough detail for the Zyur team to understand the issue. Do not send confidential text or personal information that is unnecessary for the request.