Terms and privacy
Plain-language draft. A qualified attorney will review this; until then it states our actual practice.
Terms of service
What TemplateBreaker is for
TemplateBreaker exists to improve your own writing: to show you which sentences read templated and why, and to rewrite them in your own voice. It is a writing and editing tool.
What it must not be used for
You may not use TemplateBreaker to misrepresent the authorship of work in academic, professional, or contractual settings. That includes submitting someone else's work (human or machine) as your own, and includes any use that violates your institution's academic-integrity policy. Academic-integrity offices and educators auditing sets fairly are intended customers of this product, not an adversary; we build for them. Accounts used to launder authorship will be terminated.
Our administrators may suspend an account that breaks these terms. A suspension takes effect immediately: the account is signed out of the app and its API keys stop working until the suspension is lifted. Suspension does not cancel a paid plan on its own.
What we promise, and what we do not
We promise an explanation of why text reads templated and a rewrite toward your own voice. We do not promise, and you must not rely on TemplateBreaker for, any score from any third-party AI detector. Detector outcomes are outside our control and outside this product's purpose.
Subscriptions, cancellation and refunds
Paid plans and renewal
Lite, Pro, Max and Ultra are subscriptions billed through Stripe, monthly or yearly, at the price listed on the pricing page when you subscribe. A monthly plan renews each month and a yearly plan renews each year at that listed price until you cancel. Payment details go to Stripe and never touch our servers.
Cancelling
You can cancel at any time from Billing in your account, which opens the Stripe customer portal. Your plan stays active until the end of the period you have already paid for, with its full allowances, and then your workspace drops to Starter. Nothing further is charged after that.
Refunds and proration
We do not refund or prorate the unused part of a period you have already paid for, monthly or yearly, except where the law requires a refund. Cancelling stops the next charge; it does not return the current one. If you believe you were charged in error, contact support before disputing the charge with your card issuer, and you will get a straight answer.
Prepaid API credit
API credit is a prepaid balance. Each API key is billed by your API credit. Each connected app is billed either by your subscription's monthly Humanize runs or by your API credit, as you choose. Runs billed to your subscription count toward that plan's allowances and use no credit, and a connected app billed this way can be used only through that app. Credit does not expire and is not refundable, except where the law requires a refund or the owners issue one through Stripe; Recent activity on the Credit tab of your API page shows every change.
Privacy and retention
If you visit a referral link, we use a cookie lasting up to 30 days to connect a new account to the person who referred you. We keep referral attribution, commission adjustments, and payout records for program administration and accounting. Referrers see aggregate signup and earnings counts, not your email address. Stripe collects identity and bank details for referral payouts through its hosted onboarding. We store the connected account and transfer identifiers, not your bank details.
If you arrive from a campaign link or another website, we set a first-party cookie lasting up to 30 days that records the campaign tags in the link and the referring site's name, never any text you write, and if you create an account we keep that first source with it.
To keep Starter's free Humanize runs to one set per person, when you Humanize we keep a first-party cookie with a random identifier for up to 400 days, and on Starter your browser also computes a device signal from its settings and features (such as screen size, time zone, and installed fonts) and sends it with the request. We store only a hash of each (the device signal's hash is keyed with a secret), never the identifier or the signal itself, link them to your free runs and a shortened network address, and use them only to limit free runs and prevent abuse, never for advertising. A device link not seen for 90 days is deleted.
We measure our advertising with the Meta Pixel and Meta's Conversions API. The Pixel sets the _fbp and _fbc cookies on this site and reports page views, a check on the public checker and a finished signup to Meta. From our servers we send Meta a hashed version of your email address and account id, your IP address and browser user agent as they were when you signed up or started a purchase, and for a purchase its pre-tax value and the plan bought. Never any text you write, and never a score. Meta uses this to measure which ads led to signups and purchases. If your browser sends the Global Privacy Control or Do Not Track signal, the Pixel does not load and our servers send Meta nothing about your signup or purchase. You can also manage ad settings in your Meta account. We also measure our ads in ChatGPT with the OpenAI Measurement Pixel and OpenAI's Conversions API. The OpenAI Pixel sets the __oppref cookie (the ad click id from the link you arrived from, kept up to 30 days) and the __obref cookie (a random browser reference, kept up to a year) on this site, and reports page views, a check on the public checker and a finished signup to OpenAI. OpenAI's automatic advanced matching, which we cannot switch off, also lets the Pixel pick up contact details that appear on a page, such as an email address, including one inside a draft you paste, and send OpenAI only a one-way hash of them, never the details themselves, to match ad clicks to signups. From our servers we send OpenAI a hashed version of your email address and account id, those two cookie values, and your IP address and browser user agent as they were when you signed up or started a purchase, and for a purchase its pre-tax amount, currency and the plan bought. Never any text you write, and never a score. OpenAI uses this to measure which ads led to signups and purchases. If your browser sends the Global Privacy Control or Do Not Track signal, the OpenAI Pixel does not load and our servers send OpenAI nothing about your signup or purchase. We use PostHog for product analytics on our marketing pages and in the app. Your browser sends PostHog the pages you view and when you leave them (page addresses with query strings removed), the address of the page that referred you, clicks on our marketing pages, your browser, operating system and device type, and your IP address, from which PostHog derives an approximate location. PostHog keeps a cookie and a local storage entry in your browser so repeat visits count as one visitor. Our servers also send three account milestones: when you sign up (with the campaign tags in the link you arrived from, such as utm_source, and the referring site), when you start a checkout (the plan, billing period and the offer you came from), and when you pay (the plan and the pre-tax amount). After you sign in we link these to a coded account number, never your email, and we never send text you write or a score. On the free checker, your browser also tells PostHog when you open it (with the campaign tags or referring site you arrived from, and whether you came with a draft from our homepage) and each time you run a check (whether it used the sample text), never the text itself. PostHog also records sessions on our marketing pages and in the app: the page layout, taps, clicks and scrolling, so we can see where the site is hard to use. Recordings never include what you type in a field or the writing area, and in the app every letter on a page that can show your writing is replaced before it leaves your browser. Heatmaps and click capture are off in the app. In the app, Global Privacy Control or Do Not Track turns PostHog off and our servers send it nothing about you; our marketing pages honor Do Not Track.
The free deterministic analysis runs entirely in your browser: it is free (using the analyzer and batch audit requires a free account; the checker at /check needs none). Text is not uploaded for the browser analysis itself. Saved history is on by default and you can turn it off at any time. Requesting explanations or a Humanize rewrite is optional. These send text to our servers as described below:
- Account details. We store your name, email, where you heard about us, and the date and version of the terms and privacy policy you accept. These help us run your account and understand how people find us.
- History is on by default and kept for 90 days. The analyzer runs the deterministic engine in your browser; saved history keeps your completed results in your account unless you turn it off, and an explanation request sends the text to a model provider as described in the next point.
- Requested explanations and Humanize rewrites use model providers. Requesting an explanation sends the submitted text to Anthropic (Claude). A Humanize rewrite sends the text to Anthropic (Claude) to plan the rewrite and to check that each draft keeps your meaning, and to a writer model we host ourselves on a rented GPU server (RunPod), which writes the drafts. When the Anthropic route is not configured, the explanation, planning and meaning-check text goes through OpenRouter to a model provider instead; a request that fails at Anthropic is not resent elsewhere. The browser analysis itself stays on your device.
- Saved scans expire after 90 days. Unless you turn Save my history off, completed Analyzer and Humanize results are saved to your account, including the original text, rewrites, review notes, and the version last copied. Each scan expires 90 days after it is first saved. Opening or copying it does not extend that period. You can delete a scan at any time or turn history off at any time to stop future saves. Existing scans retain their original expiry when history is turned off.
- Batch raw text is purged on completion. When a batch audit finishes, the uploaded documents are deleted; only a per-document hash and the audit report survive.
- Usage events are kept for 12 months for billing and quota accounting. They record counts, models, and costs, not your text.
- Activity and location are kept for 90 days. While you are signed in, we record which app pages you open (the page address only, never what follows it) and which features you use (analyze, humanize, batch, copy, and export). For a feature we keep counts and settings such as the number of words or documents, which result you copied, or the export format, never the text itself. With each record we keep an approximate location (country, region, and city worked out from your network address), a shortened network address with its last part removed, your browser and its main version number, your operating system, your type of device, and a random identifier for the browser tab you used. We use this to keep accounts secure, prevent abuse, help you when you contact support, and understand how the product is used. We do not record this for visitors who are not signed in, on admin pages, on sign-in pages, or for API key requests. Records are deleted after 90 days. Your account summary shows your latest page, location, shortened address, browser, and device; 90 days after your last visit all of that is cleared, and only the dates of your first and last visits and the number of pages and features you used remain. Only authorized administrators can see them.
- Request account deletion through support. Saved text, voice samples, and workspace data are removed as part of account deletion. Referral commissions and payouts require manual reconciliation before an account can be deleted; accounting records may need to be retained. You can delete individual saved scans immediately from Saved scans.
API requests process submitted text on our servers. Model-backed explanations and rewrites also send text to the providers named above. Rough drafts run on our servers only and reach no model provider. API results and Rough drafts are not automatically added to Saved scans.
Transport is TLS, data at rest is encrypted, payment details go to Stripe and never touch our servers, and webhook signatures are verified. Questions about any of this: ask, and you will get a straight answer.