Privacy policy

Effective: September 16, 2026

This policy explains how Ankush Chowdhury (“Humanizer AI,” “we” or “us”) handles information in the Humanizer AI app and support website. Cloud processing and Apple Ads attribution below apply to version 1.1 and later. The original public version 1.0 build 3 processes writing and style checks on the device and does not send app analytics or attribution. Local storage, sharing and support provisions apply to both.

Your writing at a glance

Your writing is used to complete the rewriting, quality checks and AI detection you request, then return the results to your app. Connections between the app and our service, and between our service and OpenAI, use HTTPS/TLS encryption in transit.

We do not save submitted drafts or rewrites in our backend databases or application logs, collect them into a writing dataset, or use them to train or fine-tune our models. Optional drafts and history are saved on your device. Separate usage analytics contain no writing. OpenAI's processing and retention practices are explained below.

Writing and cloud processing

When you tap Humanize, your submitted text, selected settings and frozen phrases pass through our application server on Railway to OpenAI for rewriting. OpenAI also analyzes the original and candidate rewrites to provide AI detection. Analyze sends the text you select through the same backend to OpenAI for detection. These features require an internet connection. Editing, autosave and viewing saved writing do not by themselves upload text.

Our application server handles writing temporarily in memory to fulfill the request and sends the result back to your app. We do not sell submitted writing or use it for advertising. OpenAI states that API content is not used for model training by default. Our requests use store: false to disable saved API responses. This is not a zero-retention guarantee: OpenAI may retain content for abuse monitoring (up to 30 days by default, subject to the exceptions in its policy) and temporary technical caching. Railway and OpenAI may process connection information and technical logs under their service policies. See OpenAI’s API data controls at https://developers.openai.com/api/docs/guides/your-data and Railway’s privacy policy at https://railway.com/legal/privacy.

Only submit text you have permission to share with these providers. Submitted writing may contain personal information if you include it. Humanize or Analyze starts processing; the app does not have a separate cloud-permission switch. Cancel stops waiting for a result but cannot undo data already received by a provider.

Detector cache and service protection

The backend stores keyed text fingerprints, detector versions and numerical document/sentence scores without the writing itself. The cache holds at most 10,000 results. New results remove the oldest inserted results when the cap is reached; reading an existing result does not extend its retention. An already-started detection can finish after cancellation.

To protect capacity, our server keeps request counters keyed by a hash of a client network address that changes each UTC day. The database does not store raw IP addresses; prior-day counters are replaced when the next day’s first request arrives. Hosting infrastructure may retain network information under its own policies. Numerical service capacity and failure aggregates contain no writing or client identifiers and are retained for up to 30 days.

Local storage and your controls

If enabled in Settings, the app stores the current draft and up to 50 rewrites in its local sandbox using AsyncStorage, without app-layer encryption. Turning autosave off removes the saved draft. Turning history saving off stops future saves but leaves existing history. Delete all writing data clears saved writing, the open editor, result and protected phrases. A storage warning appears if a change cannot be saved.

Uninstalling clears local app data, subject to your device backup and restore settings. Backups may retain app data under Apple’s policies. Deleting writing does not delete separate usage analytics or previously processed provider data. Contact humanizeraiteam@gmail.com for privacy requests; we will explain what we can identify and delete without asking you to send your writing.

First-party usage analytics

The app sends feature-use events, app/session openings, onboarding activity, completed/failed/cancelled operations, input-length bands, selected settings, copy/share actions, review-request attempts, operation times, app version and platform to our Railway backend. Random installation, session and operation identifiers let us count usage and repeat visits. Events never contain drafts, rewrites, frozen phrases, clipboard contents, arbitrary error messages or IDFA. The identifiers are not account credentials.

Events are batched separately from writing requests, deduplicated and retained for up to 90 days, with a 300,000-event storage cap. The local retry queue holds at most 100 events for up to seven days. Removing the app clears local identifiers, subject to backup restoration. Our password-protected dashboard shows aggregate reports. Review counters record requests to Apple’s review API, not ratings or whether a review was submitted.

Apple Ads attribution

On iOS, Apple’s AdServices framework provides a temporary attribution token. The app sends it and its random installation identifier to our backend, which exchanges the token with Apple. Tokens stay in memory for the exchange and are not stored or logged by our app or application server. The app retries temporary failures a limited number of times without blocking writing.

We retain Apple’s attribution result and available campaign, ad group, keyword, ad, organization, placement, conversion-type and App Store country/region metadata linked to that installation identifier for up to 90 days, capped at 100,000 installation records. The country/region is attribution metadata, not GPS location. This lets us compare successful rewrites, copy/share actions and return visits by acquisition source. Unavailable attribution is recorded as unknown in aggregate reporting; a false attribution response is not treated as proof of an organic download.

We do not access IDFA, request tracking permission, share usage events back into Apple Ads bidding, retarget people, sell data, use a data broker or combine activity across other developers’ apps and websites. The app does not display third-party ads or include a third-party advertising/measurement SDK. Apple processes its attribution service under its own privacy practices.

Clipboard and sharing

The app accesses clipboard content only when you invoke Paste or Copy. Clipboard content may be available to other apps or sync through Apple services. Share opens the system share sheet; your chosen destination then handles the selected text under its own privacy practices.

Support and this website

If you email support, we receive your email address and what you include. We use it to respond and resolve the request. Do not send private drafts unless needed. This website has no app-owned analytics or contact form; its hosting providers may process connection information and technical logs to deliver and secure it.

Children

Humanizer AI is a general-audience writing utility and is not directed to children below the minimum age to consent to data processing in their jurisdiction. Do not submit writing unless you are permitted to share it with the providers described above.

Changes and contact

We may update this policy as the app changes. The effective date identifies the current policy. Questions and privacy requests: humanizeraiteam@gmail.com.