Margo is an AI
Margo is an artificial intelligence, not a person. Everything Margo writes or says to you is generated by AI. This page says where we tell you so, how it is marked, and what Margo does not do.
What is generated by AI
- Everything Margo says. Replies, check-ins, suggestions, plans, summaries of your day, the drafts it writes for you, and its voice — in the app and on a phone call — are generated by AI. No person reads or writes your conversations with Margo.
- What Margo concludes about you. Margo forms views about your habits and follow-through from what you tell it and the sources you connect. They are AI inferences, shown with the evidence, and you can correct them.
- Much of this website and the apps themselves. Margo LLC uses AI heavily to build Margo: most of the text on this site — including these legal pages — and much of the apps’ code were drafted with AI and then reviewed by Margo LLC, which is responsible for them.
Who is behind it
Margo LLC is the provider of the Margo AI system and is responsible for it. Margo is built on Anthropic’s Claude models, reached through Margo’s own AI service. Spoken answers are produced by ElevenLabs or OpenAI; phone calls, if you switch them on, run through Twilio. The privacy notice lists every provider and what each receives.
Where Margo tells you it is an AI
- Before your first conversation, a notice you confirm: Margo is an AI, not a person, and can make mistakes.
- In every conversation, an “AI” badge beside Margo’s name, and a note by the message field.
- Notifications from Margo are marked “Margo (AI)”.
- In voice mode, a banner with the AI badge stays on screen for the whole session.
- A phone call from Margo starts, before anything else, by saying it is an automated AI voice and not a person.
- Drafts Margo writes for you are labelled as AI drafts, and nothing is sent until you press Send.
- If you ask Margo whether it is a person, it tells you it is an AI.
How AI output is marked for machines
- Every answer Margo’s service sends to the apps carries an `AI-Generated: true` header and a provenance field naming Margo LLC as the system’s provider.
- Margo passes the model’s text on unchanged, so marking that Anthropic applies to Claude’s output is kept. Anthropic has committed to the EU Code of Practice on marking AI-generated content.
- For spoken audio we rely on the marking our voice providers apply to generated speech, and we are confirming in writing which of their services carry it. A phone line can strip such marks, which is why every call also opens with a spoken disclosure.
- Please do not remove or alter these markings when you reuse what Margo produced.
What Margo does not do
- It does not read emotions from your face or the tone of your voice, and it does not categorise you by biometric data.
- It does not make decisions about you for employers, schools, insurers or lenders.
- It is not therapy, medical, legal or financial advice, and it is not an emergency service.
- It does not use your conversations to advertise to you, and your data is not sold.
AI can be wrong
Margo works from what it knows about you, which is never everything, and AI can state wrong things confidently. Check what matters before you act on it, and do not let Margo replace your own judgement or the people who know you. Corrections you give Margo outrank what it inferred.
Questions
Ask help@margo.global about how Margo uses AI, or to check whether something came from Margo. The rules this page follows include Article 50 of the EU AI Act and the Commission’s guidelines on it; Margo LLC aligns with, but has not signed, the EU Code of Practice on marking AI-generated content.