Rewrite existing AI text, inspect sentence-level detection signals, and edit in a browser or mobile app. Developer API plans provide separate word-based access.
Used this tool? Rate it
Used this tool? Rate it
WriteHuman AI (also branded WriteHuman) is a WriteHuman LLC service for rewriting existing text, AI text and image detection, grammar suggestions, and submission history. It runs through a website, mobile apps, a browser extension, and developer services.
The company describes its AI humanizer as changing cadence, vocabulary, and sentence structure while preserving meaning. It distinguishes this from generating an original draft or removing plagiarism: the rewriter cannot supply ideas, sources, or arguments absent from the input.
WriteHuman AI describes replacing formulaic filler, varying repetitive sentence structures, mixing sentence lengths, and reducing overused transitions. These are vendor-described changes to expression, rather than independent measures of accuracy or quality.
The mobile app offers Standard, Professional, Blog, Casual, Creative, Scientific, Technical, and Academic tones, plus an Enhanced toggle for a stronger rewrite. Web plan limits also distinguish output variations from request length and monthly allowance.
The homepage AI detector demonstration assigns sentence probabilities: under 40% is labelled as reading human, 40–60% as mixed, and over 60% as flagged as AI. These displayed interpretation bands are not proof of authorship.
The API provides humanization, detection with an AI/human/mixed verdict and optional sentence breakdown, and a read-only usage lookup. The usage endpoint reports remaining words, plan, and top-ups without consuming allowance.
Premium API processing generates and scores three rewrites, returning the highest-scoring version and two alternatives. All Premium sizes include built-in detection, detection-aware processing, and priority processing. Ranking reflects the service’s assessment, not a guaranteed external detector result.
The extension combines a grammar checker and humanizer inside the field being edited:
WriteHuman AI is positioned for revising articles, blog posts, marketing copy, emails, landing pages, newsletters, and reports, from snippets to longer supplied pieces. These are vendor-described applications, not measured improvements in search ranking, engagement, or acceptance.
The company names founders and product teams revising landing pages, launch posts, and changelogs, and bloggers or creators maintaining a consistent voice across pieces. The task remains revising supplied material rather than generating missing product facts.
Developers processing their users’ text are responsible for the necessary rights, notices, consents, and end-user support.
The API requires an account and active API subscription, billed and cancelled separately from the website plan. The terms name Pro, Ultra, and API subscribers for MCP access. Personal MCP clients exclude shared connectors, multi-user bots, and team accounts; those use the API.
Website prices default to USD, with applicable taxes additional. Monthly rates and annual equivalents are separately displayed; annual cards state the yearly charge and a “4 months free” label.
| Web plan | Monthly billing | Annual equivalent and charge | Humanizations per month | Maximum input per request | Output variations |
|---|---|---|---|---|---|
| Basic | $20/month | $12/month; $144 billed yearly | 80 | 600 words | 2 |
| Pro | $29/month | $19/month; $228 billed yearly | 200 | 1,200 words | 3 |
| Ultra | $59/month | $39/month; $468 billed yearly | Marketed as unlimited | 3,000 words | 5 |
Web Free lists three monthly humanizations, 250 words per request, one output, ten detection checks, and ten image scans. Its quality row is Basic; paid tiers are Enhanced.
Ultra lists 1,000 MCP requests/month. Its advertised unlimited writing remains subject to the legal caps below.
Mobile starts without an account, email, or card, with separate allowances of five humanizations, five text detections, and five image detections. Each refills on a rolling 30-day window. This is the mobile offer, distinct from the web Free table.
API plans are monthly and separate from consumer subscriptions:
| API plan | USD per month | Included words per month | Words per call | Outputs per request | Requests per minute |
|---|---|---|---|---|---|
| Standard | $29 | 125,000 | 2,000 | 1 | 40 |
| Premium 400K | $69 | 400,000 | 4,000 | 3 | 120 |
| Premium 800K | $129 | 800,000 | 6,000 | 3 | 160 |
| Premium 1.5M | $225 | 1,500,000 | 6,000 | 3 | 160 |
| Premium 3M | $419 | 3,000,000 | 6,000 | 3 | 160 |
Top-up packs cost $25 for 100,000 words, with up to ten packs per purchase. Monthly included words reset without rollover. Top-ups follow included words and never expire, but purchase and use require an active API subscription. After cancellation, the balance remains inaccessible until an API plan is active again. The current ceiling is 1,000,000 top-up words. They are nonrefundable, nontransferable, and forfeited on account deletion or termination.
Optional auto top-up charges a chosen amount below a selected remaining-word threshold, within a chosen 30-day charge limit. Disabling it stops future charges without refunding past ones. The provider may pause it after payment failure, refund, chargeback, or payment-method change and require reconfirmation.
Paid subscriptions automatically renew for the same period at the then-current price plus tax. Website cancellation is in My Account; API cancellation is in its dashboard. Stripe cancellation takes effect at period end, retaining paid access until then before moving to Free. Uninstalling or nonuse does not cancel.
Website plan changes are immediate: upgrades charge the prorated difference, while downgrades credit unused time to future invoices rather than cash. A pause may be offered, stopping charges and paid access, with some tools potentially unavailable entirely. Billing auto-resumes on the displayed date unless cancelled; early resumption or changing plan ends the pause.
Payments are nonrefundable except as legally required. Exclusions include partial periods, unused time or resources, top-ups, nonuse, suspension, termination, and downgrades except the account credit. A discretionary refund creates no obligation to repeat it. Trial or introductory offers show their end and subsequent charge, payable unless cancelled before that period ends; promotions renew at the regular price.
Apple and Google handle store subscription billing, renewal, cancellation, and refunds under their policies, which prevail over conflicting in-app billing terms. Apple auto-renew must be disabled at least 24 hours before period end; it charges renewal in the preceding 24 hours. Google sets its own timing.
Grammarly’s Paraphraser describes free-account rewriting, six preset styles, and custom tone and style direction applied to later rewrites. These are vendor-stated controls, rather than measured advantages over WriteHuman AI.
Its Paraphraser lists English, Spanish, French, German, Portuguese, and Italian. Style controls and specific language coverage provide a comparison with WriteHuman AI’s broader advertised website coverage and English-only extension checking.
QuillBot offers another rewriting route: its free Paraphraser includes Standard and Fluency modes with limited usage. Its Synonym Slider adjusts how much vocabulary changes, and clicking a word shows replacement choices. These controls target the extent of an edit, rather than establishing any detector outcome.
WriteHuman AI markets text as indistinguishable from human writing and promotes detector-passing scores. Its terms do not guarantee any detector score, label, or result, including its own. Detection is probabilistic, can be wrong, and is not proof of authorship or a permitted sole basis for decisions about a person, including employment or discipline.
Marketing describes preserving meaning, while the terms say outputs may be inaccurate, incomplete, or change it. The promoted benchmark is published by WriteHuman itself, so its rankings are vendor-published evidence rather than independent validation.
The terms prohibit academic dishonesty and breaching applicable employer, publisher, or platform rules on AI use or authorship. An Academic tone does not remove those restrictions.
“Unlimited” means no stated monthly allowance for a feature, but the terms retain tool-dependent monthly request caps, per-request word limits, rate limits, and fair-use controls. Requests can be refused until the next usage period after a cap is reached. Excess use can be throttled, restricted, or suspended.
Humanizer and detector text and results are saved in history; uploaded images are not saved to accounts or databases. Processing uses company models and AI providers. Text samples may be shared with third-party detection services for testing without a user’s name or email.
Consumer website, app, and extension text, outputs including additional versions, feedback, and ratings train, evaluate, and improve models. There is no opt-out, and content already used in training cannot be removed from the model. Staff or contractors may review it, subject to the extension restrictions below.
API and MCP text is not used for training or stored in databases. It is processed for results, abuse screening, and model reliability checks; service logs may temporarily contain request data and check results during limited testing periods. Metadata supports billing, limits, security, and support. MCP connection and tool-call records exclude text.
The extension reads active multiline or rich-text fields, potentially including personal communications, not single-line password, email, payment, or search fields or the rest of the page. Grammar and spelling text is processed without storage. Sentence suggestions may be cached for up to seven days without an account link; humanization and detection history retains the training-data exception.
Staff and contractors read extension text only with specific consent, for security, for legal compliance, or in aggregated and anonymized internal operations. Anonymous quality signals may include flagged fragments, chosen replacements, and site domains without account or install identifiers.
History lasts until deletion of it or the account, but training/evaluation copies can remain while useful for model development. Signed-out content can remain for operations, improvement, and training. Billing retention follows tax/accounting/legal needs, support follows request handling, and device, usage, security, error, and marketing records follow their stated purposes. Deletion can leave legal, billing, security, fraud-prevention, suppression, de-identified, training, and replacement-cycle backup records. Previously shared information follows third-party retention policies.
Privacy rights depend on residence and applicable law, subject to legal limits and exceptions. They can include access, copying, correction, deletion, objection, restriction, consent withdrawal, and advertising opt-outs. Access, deletion, and correction require identity verification and may be denied if unverifiable; sale/sharing opt-outs do not.
The company disclaims selling personal information for money but says advertising may count as sale/sharing under state law. Browser/account opt-outs and Global Privacy Control apply to website advertising. They do not delete earlier data or stop necessary cookies, own analytics/error reporting, affiliate tracking, or emails. Apps use separate opt-out routes.
Users retain existing input rights but grant a worldwide, perpetual, irrevocable, transferable and sublicensable license for operations, improvement, model development, and other stated purposes. It survives deletion and cancellation, subject to non-waivable privacy rights and developer/extension exceptions.
WriteHuman assigns any rights it holds in delivered outputs to users, subject to the terms and content license, allowing lawful commercial use. It promises neither originality, accuracy, copyright protection, nor freedom from third-party rights. Similar outputs delivered to others are not the first recipient’s property.
Both humanization and detection consume input words. Chinese, Japanese, Vietnamese, and character-heavy input such as unspaced text, links, or code may use a character-based or other estimate. The provider’s count controls, without retroactive usage adjustments, credits, or refunds, including code errors, loops, or misuse of keys.
Deletion immediately cancels Stripe website/API plans, including paused or past-due plans, forfeiting unused resources and paid time without refund. It does not cancel Apple or Google subscriptions, which continue until cancelled through the store. Data retention exceptions still apply.
