Privacy

What Fol sends, what it keeps, and what it never sees. Last updated 18 September 2026.

Fol turns speech into text. You hold a key or tap a button, talk, and the words appear where you were typing. This page describes exactly what leaves your device to make that work.

What we collect

Audio, while you are dictating

When you dictate, the recording is sent to our servers, transcribed, and the text is returned to your device. The audio is not retained — it exists only for the moments it takes to transcribe it. It is not used to train models and it is not shared with anyone beyond the speech provider that performs the transcription.

Word counts

We record how many words you dictated, when, and which of your devices it came from. This is what your weekly allowance is measured against. It is a number — it does not include what you said.

Your account

If you sign in, we store your email address, your plan, and the sign-in tokens that keep you signed in. Signing in with Google gives us your email address and nothing else.

The name of the app you dictated into

Sent with each dictation, so Fol can match how you write there and your history can show that a line went into Gmail rather than Slack. We store the app's name alongside your word count, never anything in the app. On Android this name comes from the accessibility service.

Your dictionary

Words you add to your dictionary — names, product terms — are stored with your account so every device you use spells them the same way.

What we do not collect

• Your transcripts. By default, what you dictate is stored only on your own device and is never sent to us for storage. We never log transcript text anywhere. The one exception is a switch you have to turn on yourself — “Help train Fol”, below — and even then what is kept carries no account.
• Your screen. On Android, Fol uses the accessibility service to find the text field you are typing in and to insert your words at the cursor. It reads that one field and nothing else on screen, and it does not record or transmit the contents of your screen. See “The accessibility service” below.
• Your contacts, photos, location, or browsing. Fol never asks for these.
• Advertising identifiers. Fol carries no advertising and no third-party trackers.

Things you can turn on, which are off unless you do

Each of these is a deliberate choice in Settings, and each is off until you make it.

• History syncing. Off by default: your dictation history lives on your device only. Turning it on mirrors your lines to your account so a second device starts with the same list. You can turn it off again and delete what was synced.
• Help improve recognition. Sends word pairs — a word you corrected, against what Fol heard — with no account attached, so Fol mishears that word less often. Words only: anything longer than three words or forty characters is refused.
• Anonymous usage statistics. Counts of a short, fixed list of in-app events, with no account attached and no content.
• Help train Fol. The only setting that keeps whole sentences. With it on, each dictation contributes the pair Fol’s writing is judged on: what the speech model heard, and what Fol wrote from it. Stored with no account attached and no app name, so it cannot be traced back to you. Never your audio. Nothing is sent when Fol changed nothing, and nothing is sent at all while “Keep everything on this device” is on. We use it to measure and improve how Fol cleans up speech — there is no honest way to do that against examples we wrote ourselves.
• Only my voice. A few seconds of your voice, recorded once in the app and kept on your device. While it is on, the sample is sent with each dictation so the transcription keeps only your words and ignores other people talking around you. It is passed to the speech provider for that dictation and never stored on our servers. Turning it off stops sending it; deleting the sample, or signing out, removes it from your device.

Four further features send the specific text you point at, only at the moment you ask for them: a rewrite, a question to Ask Fol about text you selected (on Android, the whole field when nothing is selected), a writing sample you paste in to teach Fol your style, and telling Fol that a dictation “doesn’t sound like me”. None of that text is stored — only a short description of your writing style, which never contains what you wrote. Reading a line aloud in Fol’s voice works the same way: that one line is sent to be turned into speech, and neither the text nor the audio is kept.

The accessibility service

Fol’s Android app uses Android’s accessibility API for one purpose: to put your dictated words into the app you are typing in. Fol is not an accessibility tool for people with disabilities, and it asks for this permission only after explaining, inside the app, what follows here.

• What it reads. The text field you are typing in: where it is on screen, and the text already in it, so your words go in at the cursor without replacing what you wrote. To find that field it looks for text boxes on the screen; it does not read what anything else on the screen says.
• What leaves your device. The name of the app you are dictating into, sent with each dictation as described above. The text of the field stays on your phone, except when you choose Ask Fol: then what you selected, or the whole field when nothing is selected, is sent with your question so Fol can answer it, and is not kept.
• What it never does. It does not watch what you type, record your screen, tap or act on your behalf, or change your settings. Nothing gathered through it is used for advertising, sold, or shared.

Permissions, and why each one exists

• Microphone — to hear you when you dictate. Fol records only while you are actively dictating.
• Accessibility service — to insert your words into the field you are typing in, as described above.
• Display over other apps — to show the small Fol button beside your keyboard.
• Notifications — to tell you when something needs your attention, such as Android switching the service off.

Who else sees your data

We use a small number of service providers to run Fol, each only for its own job: OpenAI performs the transcription and the read-aloud voice, Railway hosts our servers and database, and Lemon Squeezy processes payments if you subscribe on our website. They act on our behalf and may not use your data for anything else. We do not sell your data, and we do not share it for advertising.

How long we keep things

Audio is never kept. Word counts and your account details are kept while your account exists. Synced history, if you turned that on, is kept until you delete it or delete your account.

Your choices

You can delete individual dictations, turn syncing off, or delete your account and everything attached to it at any time. See Delete your account.

Children

Fol is not directed at children under 13 and we do not knowingly collect their data.

Changes

If this policy changes in a way that affects what we collect, we will say so in the app before it takes effect.

Contact

Questions, requests, or anything you want deleted: ad@coach-bot.de.