Responsible AI use policy

AI writing tools can be genuinely useful. They can help you get past a blank page, tighten a clumsy paragraph, or write in a second language with more confidence. They can also be misused. This policy explains where we draw the line. It forms part of our terms of use.

Good ways to use Text Humanize

• Editing your own drafts, or AI-assisted drafts, to make them clearer and more natural
• Adapting tone, for example making an email more friendly or a report more formal
• Improving business writing, marketing copy, help articles and personal correspondence
• Getting an alternative phrasing to compare with your own when English isn’t your first language
• Studying how wording changes affect clarity, using the “Show changes” comparison
• Coursework, when your instructor or institution allows AI editing tools

Uses we don’t allow

• Submitting AI-generated or AI-rewritten work as entirely your own in a course, exam or application that doesn’t permit it
• Rewriting someone else’s work to hide plagiarism or avoid crediting the source
• Producing fake reviews, testimonials, spam, scams or coordinated inauthentic content
• Creating misinformation or impersonating real people or organisations
• Disguising content to get around platform rules or moderation systems
• Any unlawful, harassing, hateful or harmful purpose

A note for students

Schools set very different rules. Some encourage AI for brainstorming and editing; others prohibit it for graded work entirely. Before using any AI tool on an assignment:

1. Read the policy in your syllabus or your institution’s academic integrity guidance.
2. Ask your instructor if it isn’t clear whether editing tools are allowed.
3. Disclose your use when disclosure is required, and describe what the tool did.
4. Keep your drafts and notes. They show your thinking and your writing process.

Our guide to using AI writing tools responsibly as a student goes into more detail.

About AI detectors

We don’t design Text Humanize to defeat AI detectors, and we don’t promise any detector result. Detectors are unreliable: they flag human writing by mistake and miss AI writing. If you’re worried about being wrongly flagged, the best protection is an honest process and a record of your drafts, not a rewriting tool. See what to do about AI detector false positives.

Check before you share

You remain the author and you’re responsible for what you publish or submit. Always:

• check facts, numbers, names and quotations against your sources;
• make sure the rewrite still says what you mean;
• credit sources and follow citation rules (see how to cite AI tools);
• add your own knowledge, examples and judgement. That’s what makes writing worth reading.

How this policy applies in practice

Academic work

Universities and schools have published a wide range of positions on AI writing tools. Some institutions, such as those following the European Council for Applied Linguistics guidance, allow AI editing for clarity as long as disclosure is made. Others, like many US universities that updated their honor codes after ChatGPT’s release, prohibit AI-generated text entirely for graded assignments. The safest approach is to assume disclosure is required until you have confirmed otherwise in writing.

When you do use an editing tool on coursework, keep a short record: the original draft, the tool you used, what you asked it to do, and your final edit. This log demonstrates process, which is what instructors actually assess. It is far more useful than trying to explain after the fact that a passage “sounds too good”.

Workplace and professional writing

In business settings, the risk is not usually academic misconduct — it is accuracy. AI rewriting can subtly change meaning, drop caveats, or turn a qualified statement into an overconfident one. Before sending anything rewritten, re-read the passage against your source material. Pay particular attention to:

• Negations. “We cannot guarantee” can become “We guarantee” after a careless rewrite.
• Numbers and dates. Round figures and approximate dates are frequently altered.
• Attribution. A claim that was cited to a source may lose its attribution.
• Tone calibration. A tool optimised for friendliness may undercut the seriousness a legal or HR communication needs.

Public content, reviews and marketing

Producing fake reviews, inflated testimonials, or undisclosed AI-generated endorsements is prohibited by consumer protection law in many jurisdictions and by the platform terms of virtually every marketplace and social network. Rewriting a genuine review to make it more persuasive is still misrepresentation if it changes the meaning. The same applies to sponsored content that presents itself as organic opinion.

For marketing copy, the practical standard is simple: would you be comfortable if a customer read this text next to a note saying “written with AI assistance”? If the answer is no, the copy needs human input, not just a rewrite.

Disclosure: when and how

If your institution, employer, or audience expects disclosure, be specific rather than vague. “This was written with AI help” is weaker than “I drafted this myself and used an AI tool to tighten the phrasing of the conclusion.” Naming what the tool did shows you were in control of the process.

For published content — blog posts, reports, articles — a short disclosure line is usually enough. Something like “Edited with AI assistance; all facts verified by the author” communicates both the tool use and your responsibility for accuracy.

Why detectors are not the answer

We don’t design Text Humanize to defeat AI detectors, and we don’t promise any detector result. Detectors are unreliable: they flag human writing by mistake and miss AI writing. Published research on detector accuracy has found false-positive rates as high as 61% for non-native English writers in controlled studies, and even well-known commercial detectors have been shown to misclassify excerpts from the Bible and passages written before generative AI existed.

If you’re worried about being wrongly flagged, the best protection is an honest process and a record of your drafts, not a rewriting tool. See what to do about AI detector false positives.

Reporting misuse

If you believe Text Humanize is being used in violation of this policy, email hello@text-humanize.com with as much detail as you can: what you saw, where, and why you believe it constitutes misuse. We review every report. Because we don’t require accounts or store submitted text, we generally cannot act on reports that concern a specific piece of text after the fact — but patterns of abuse reported by multiple people help us improve our technical safeguards.

Enforcement

We may limit or block access where we see clear misuse, such as automated abuse. Because we don’t require accounts or read submitted text, we rely on users to act responsibly. Enforcement decisions consider intent, scale and repetition: a one-off mistake is treated differently from coordinated or commercial-scale abuse. This policy may be updated as norms and regulations around AI-assisted writing continue to change; the date at the top of this page reflects the last revision.