Privacy Policy
What changed on 20 July 2026: this website now uses cookieless analytics and offers an email signup for launch notification. Both are described under This website below. The app itself is unchanged.
This policy describes the current development build. Conversa is not yet on the App Store. The build described below calls OpenAI and Azure directly using keys held on the device; the released version will instead run those calls through Conversa's own service, covered by the subscription described on the FAQ. That changes what leaves your device and who processes it, so this policy will be rewritten — naming every sub-processor, and with a new effective date — before the app is released. It will not be released under the policy below.
The short version: the Conversa app has no accounts, no ads, no tracking and no analytics, and collects nothing about you. This website counts page views without cookies or any personal data, and will store your email address if — and only if — you type it into the launch-notification box. The only data that leaves your device in the app is what's needed to power the AI conversation and speech features, sent directly from your device to OpenAI and Microsoft Azure using API keys that you supply.
What the app collects
Nothing. Conversa is built without a backend: there is no Conversa server that your data could be sent to, and the app contains no analytics or advertising SDKs.
This website
Two things on conversa.im are worth being precise about.
Analytics
This site uses Vercel Web Analytics to count page views, so I can tell which pages are useful and which are ignored. It was chosen specifically because of what it does not do:it sets no cookies, stores nothing on your device, and creates no persistent identifier that could follow you between visits or across other websites. What I see is aggregate — page, country, referrer, browser — never an individual.
This is why there is no cookie banner on this site: there are no cookies to consent to. Google Analytics was considered and rejected, because it would have meant both cookies and sending your visit to Google. Data is processed by Vercel, this site's host — Vercel's privacy policy.
The launch-notification list
If you enter your email address to be told when Conversa launches, that address and the date you submitted it are stored in a database I operate, hosted in the European Union. That is all that is stored — no name, no IP address, no browser details, nothing about which pages you read.
The address is used for exactly one thing: to email you when the app is released. It is not added to a newsletter, not used for marketing anything else, and never sold, shared or passed to anyone. Ask me to remove it at any time, at peterjduffy@gmail.com, and it's deleted — no confirmation loop and no attempt to talk you out of it. The list is deleted entirely once the launch email has gone out.
Your legal basis, if you want the formal version: consent, which you gave by typing the address in, and which you can withdraw the same way.
Your API keys
Conversa works with API keys that you create and own — an OpenAI key for the AI tutor and a Microsoft Azure Speech key for voices, voice input, and pronunciation scoring. These keys are stored in the iOS Keychain on your device and are used only to call those services directly. They are never transmitted anywhere else.
Third-party AI and speech services
When you use the app's features, some data necessarily goes to the providers you hold keys with, directly from your device:
- OpenAI — the text of your conversation with the tutor (messages you type or speak, and the conversation so far) is sent to the OpenAI API to generate replies, corrections, suggestions, and feedback reports. See the OpenAI privacy policy.
- Microsoft Azure — your voice recordings are sent to Azure Speech Services for transcription (voice input) and pronunciation scoring, and the tutor's text is sent to Azure to be turned into speech. See the Microsoft privacy statement.
Your relationship with these providers is your own — the data they receive is governed by their terms and your account settings with them, not by us.
What stays on your device
- Conversation transcripts and their end-of-chat feedback reports are saved to a local database on your device, so you can search past conversations in History and pick one up where you left off. You can delete any conversation from History. Session-time aids — corrections, word glosses and suggested replies — are not saved.
- Things the tutor remembers about you — if you enable tutor memory, short facts drawn from your conversations (for example that you're learning for a move to Lisbon) are stored on your device and included in the prompt sent to the AI provider, so the tutor stays consistent between sessions. Settings lists every remembered fact and lets you delete them individually or all at once.
- Pronunciation history (your scores and trouble sounds) is stored in a local database on your device.
- Generated audio is cached locally so repeated phrases don't need re-fetching.
None of this is sent to us — there is no "us" to send it to. Deleting the app deletes all of it.
Microphone
Conversa asks for microphone access so you can speak to the tutor and record pronunciation practice. Recordings are used only for the features described above; the app does not listen in the background.
Children
Conversa is not directed at children under 13.
Changes
If the app gains features that change any of the above (for example, optional iCloud sync of your practice history), this policy will be updated here first, with a new effective date.
Contact
Questions? Email peterjduffy@gmail.com.