What we can and can't see

What we can't see: your words

Your chats, cards, notes and what the guide remembers are locked with a key that belongs to your account. No one on our team reads them, and they're never sold or used for ads. The only exceptions are explained below.

What we can see

  • Your account: a random ID, and your email if you sign in.
  • Counts: how many chats and days, thumbs up or down, and moods and feelings only as totals across everyone, like how many people tapped Meh today. Never yours by name.
  • Your age range and country, kept with your account so we can help if you write to us.

The AI that replies

To reply, each message goes to the AI that writes it: Anthropic Claude (USA), or Google Gemini (Europe) if we switch to it. Google Vertex AI (India) uses it to find the cards that fit your chat. They don't train on it. Anthropic keeps it for up to 30 days and may check it for safety and security; how long Google may hold it is set by its own terms. That's also why this isn't end-to-end encryption: the AI has to read a message to answer it.

The exceptions

A safety flag. If our safety system flags a message as a risk of serious harm, a person on our safety team can open that one message to check you get the right help.

If you choose to share. If you turn on “Help improve Dr. Thoughts”, a small review team can read the chats you shared. It's off unless you turn it on, and turning it off stops it at once.

Each opening is logged and shows in your Access log. Both are switched off today.

Also

If you write to us in Get in touch, we read that message. Opening a full email address is logged. If you never sign in, everything is deleted after 7 days without use. Full privacy policy

To know where Dr. Thoughts is available, your country (and your state in the US) is read from your network address, which isn't kept. IP location by DB-IP (db-ip.com).

Your control