What this is for.
And what it isn't.

1. What this document is

This is the rulebook for how HumanGPT may and may not be used. It applies to every visitor on the free tier, every Pro subscriber, every Founders edition holder, and every developer hitting the API. It applies whether you submit one sentence or a 5,000-word document. By using the service you agree to these rules. If we find use that breaks them, we may rate-limit, suspend, or terminate access.

2. What HumanGPT is built for

HumanGPT exists to make AI-generated text read more like writing produced by a human. The core legitimate use cases we built it for, and the ones we are happy to support:

• Polishing your own AI-assisted first drafts before they go to a real audience.
• Smoothing the output of machine translation so it stops sounding mechanical.
• Rewriting marketing copy, product descriptions, and landing-page text that came out of an AI tool and reads as such.
• Cleaning up internal documents, meeting notes, brainstorms, and research summaries that you generated with AI to save time.
• Helping non-native English speakers (and writers in any of our 30+ supported languages) produce text that reads naturally for native readers.
• Editing AI-assisted blog posts, newsletters, and articles so they match a real human voice before publication.
• Working on creative writing, fiction, and personal essays where AI was used as a brainstorming partner and the writer wants the final draft in their own voice.
• Using the API for legitimate product features inside your own software (with end-user disclosure where appropriate).

3. What HumanGPT is NOT for

Some uses of a writing tool cross from legitimate quality work into deception, fraud, or harm. Those uses are off-limits. The list below is non-exhaustive; if a use feels close to anything below, treat it as prohibited until you have written confirmation from us.

• Submitting humanized AI text in any context where the rules of that context prohibit AI assistance. Examples: an academic institution that bans AI in coursework, a publication that bans AI-assisted submissions, a job application requiring a writing sample produced personally.
• Producing fraudulent documents of any kind: fake reviews, fake testimonials, fake academic citations, fake medical advice, fake legal advice, fake financial advice, fake credentials.
• Impersonating a real, identifiable person, dead or alive. This includes generating text in a named person's voice in a way designed to deceive readers about authorship.
• Generating content that defames, harasses, threatens, dehumanizes, or incites violence against any person or group.
• Generating sexual content involving minors. Generating non-consensual sexual content of any kind. Generating realistic content sexualizing identifiable real people without consent.
• Generating content intended to facilitate fraud, identity theft, financial scams, romance scams, phishing, social engineering, or any other deception aimed at extracting money or data from victims.
• Generating content designed to mislead voters, manipulate elections, spread organized political disinformation, or impersonate political figures or institutions.
• Generating instructions, code, or text that materially aids the creation of weapons (including chemical, biological, radiological, nuclear, or high-yield explosive weapons), the targeting of critical infrastructure, or large-scale violence.
• Generating content designed to bypass safety, moderation, or detection systems for harmful purposes (this is distinct from making writing read as natural human prose, which is the legitimate function of HumanGPT).
• Generating spam at scale: SEO doorway pages, AI-content farms, link-farm articles, scraped-then-spun content for monetization, and similar tactics that pollute the web.
• Reverse-engineering, scraping, or extracting our model behavior, prompts, or training-data signals to build a competing humanizer service.
• Reselling raw HumanGPT output as your own humanizer-as-a-service product.
• Using the service to violate any applicable law in your jurisdiction or the jurisdiction of the people your output reaches.

4. Academic integrity, professional integrity, platform rules

We get asked about this often, so we will be direct: HumanGPT is a writing-quality tool, not a permission slip. If you study at a school that bans AI in essays, our tool does not make AI use acceptable in those essays. If you write for a publication that requires disclosure of AI assistance, our output still requires that disclosure. If you submit work to a client who expects original human writing, what you submit must match what they expect.

The rules of any institution, employer, publisher, or platform you submit work to apply to your use of HumanGPT inside that context. They are between you and them. We will not pretend otherwise, and we ask users not to either.

5. AI-detection bypass and what we mean by it

We say openly that HumanGPT output reads as human across major AI-detection tools. We test against GPTZero, Turnitin, Originality.ai, Copyleaks, ZeroGPT, Sapling, Winston, and Crossplag and publish weekly results. That capability is the result of doing the writing work properly: increasing perplexity and burstiness, varying sentence rhythm, removing AI-tell vocabulary, and matching register. It is not a magic wrapper around fraud.

We do not market HumanGPT as a way to defeat platform terms or institutional honor codes. We market it as a way to write text that does not sound machine-generated, because most AI drafts sound machine-generated and that is a real writing-quality problem. Use the capability the way you would use any sharp writing tool: where it serves writing that you are entitled to publish, submit, or send.

6. Special rules for high-risk content

Some content categories require extra care because the harm of misuse is high. The following uses are prohibited even when the underlying request seems benign:

• Medical advice, prescription information, or health-condition guidance presented as authoritative or personalized. We will rewrite general medical information for readability; we will not produce text positioned as a substitute for a clinician.
• Legal advice for a specific person, case, or jurisdiction presented as authoritative. We rewrite general legal information; we do not generate documents intended to function as legal advice tailored to an individual situation.
• Financial advice for a specific person, portfolio, or transaction presented as authoritative. We rewrite general financial information; we do not generate text positioned as personalized investment, tax, or banking guidance.
• Crisis-line scripts, suicide intervention scripts, mental-health emergency text, or anything that purports to substitute for a trained crisis counselor.

7. Children and minors

HumanGPT is not directed at children. The service is intended for users 16 years of age or older. We do not knowingly collect or process content from users under 13. If you are under 16, you may not create a paid account; the free tier may be used only with the involvement and consent of a parent or guardian.

Generating sexual content involving minors is absolutely prohibited and will result in immediate termination, refund forfeiture, and where required by law, reporting to the appropriate authorities (in the United States, NCMEC). There is no edge case here.

8. Spam, scraping, and bulk abuse

The free tier is for human visitors using the tool through the website. Programmatic access goes through the API, which has its own terms, rate limits, and pricing. The following bulk uses are prohibited regardless of tier:

• Running headless browsers, scrapers, or scripts against the website to produce humanized text without an API key. The website rate limits exist for human use; circumventing them is abuse.
• Using the service to generate hundreds of similar pages or articles for the purpose of low-quality SEO at scale, doorway pages, or content farms.
• Bulk-rewriting other people's copyrighted text without permission, regardless of whether the original was AI-generated or human-written.
• Scraping the service to build a derivative model, training set, or competing product.

9. Intellectual property

You may only submit content for which you have the right to make derivative versions. That includes your own original writing, your own AI-assisted drafts, content you have licensed for derivative use, content you have permission from the author to rewrite, and content in the public domain. Submitting copyrighted text you do not have rights to is a violation of these rules and may also be a violation of copyright law in your jurisdiction. The legal liability for that submission sits with you, not with HumanGPT.

10. How we enforce this

We do not pre-screen submissions and we do not store full submission content. What we do have are signals: rate-limit patterns, abuse reports from third parties, automated flags for obvious prohibited categories at the prompt-construction layer, and direct human review when something is reported to us. When we believe a use violates this policy, our enforcement options, in rough escalation order:

• A warning email asking you to stop and explaining what we believe was violated.
• Temporary rate-limit reduction or feature suspension.
• Account suspension pending review.
• Permanent termination, with refund handling per our Refund Policy and Terms of Service.
• For severe violations (CSAM, threats of imminent harm, clearly illegal activity), immediate termination and, where legally required, reporting to authorities.

We err toward warning first when intent is unclear. We do not warn first for the categories listed in section 7 (minors), section 3 (large-scale violence, fraud, CSAM-adjacent content), or for users who have already received a prior warning.

11. Reporting abuse

If you believe HumanGPT is being misused, including content generated by HumanGPT being used to harass, defraud, impersonate, or defame, please email [email protected] with as much detail as possible: where you encountered the content, what you believe is being violated, and any evidence you can share. We treat every report seriously and respond within 48 hours during business days. Reports involving immediate safety concerns or suspected CSAM are handled within 24 hours including weekends.

12. Appeals

If your account is rate-limited, suspended, or terminated and you believe the action was a mistake, reply to the enforcement notice or email [email protected]. A real person reviews every appeal. We respond within 5 business days with either a reinstatement, a clarification, or a final answer. If your appeal is upheld and we suspended a paid account in error, we restore service and credit any lost time at no cost.

13. Changes to this policy

We update this document when the threat landscape, the regulatory landscape, or our own product changes. Material changes are announced 30 days in advance via email to active users and on the homepage. Continued use after the effective date of an updated policy constitutes acceptance. If you disagree with a change, you may cancel any paid plan before the effective date and we will refund any unused billing period in full.

14. Contact

Abuse reports: [email protected] · Appeals and legal questions: [email protected] · Everything else: [email protected]. We respond within 48 hours during business days, sooner for safety-critical issues.