An AI period tracker: what it knows, and how far to trust it

What Bloom’s AI assistant can do with your record, what it sends to Anthropic to answer you, and the limits of any AI that talks about health.

Last reviewed 15 October 2026Checked against the sources listed below

This page is for you if you are curious about a period tracker you can talk to, and want a straight answer to the two questions that matter: what the AI actually knows about you, and how far you should trust what it says. It covers what Bloom AI does, where your words go when you ask it something, and the limits of any AI that talks about health.

What “AI” means in a period tracker

Apps use the word for two different things. Some call their period predictions AI. Others mean a chat assistant you can ask questions. In Bloom they are separate:

• Predictions are not AI. Bloom works out your next period and your phases with fixed, explainable rules: the median of your last six cycles, ovulation counted back from your next period, and a late flag measured against your own variation. The same logs always give the same answer.
• Bloom AI is the assistant. It is a chat you can open from the app, plus short notes it writes for your phase and replies to your daily check-ins.

Keeping these apart matters. Your dates never depend on a language model having a good day, and you can see why a prediction says what it says. There is more on how the phases are worked out on the cycle phase page.

What Bloom AI does

• Answers from your own record. When you ask why this week feels harder, it can look at your current phase, your recent check-ins and notes, and the goals you chose, rather than giving a textbook answer about an average person.
• Drafts plans you approve. Ask for a week of easy dinners for your luteal phase, or a gentler workout plan, and it drafts one. Nothing lands in your planner until you accept it.
• Takes voice. You can hold to talk instead of typing.
• Works in every mode. In pregnancy and postpartum mode it answers with your week or recovery stage in mind.

There is a daily limit, currently 20 requests a day, including the short notes Bloom writes for you. It needs you to be signed in, and it writes in English.

Where your words go

This is the part worth reading before you type anything personal. When you ask Bloom AI something, the request goes from your phone to Bloom’s own server, and from there to Bloom’s AI provider, Anthropic, over encrypted connections. It carries what is needed to answer you: your message, and a short summary of your context such as your first name, your phase or week, your recent logs and notes, the goals and health considerations you chose, and your focuses.

Anthropic writes the reply under commercial terms that do not allow it to train AI models on that data. Bloom’s key for Anthropic stays on Bloom’s server and is never in the app on your phone. Your conversations are kept on your phone and in your account, and deleting your account deletes them with everything else. The privacy page covers the rest of what Bloom stores.

It also helps to know where the law stands. The U.S. health department says HIPAA generally does not protect health information you keep in apps on your own phone, so what protects you is the app’s own design and its privacy policy.^[1] That applies to every AI health app, not only Bloom.

How far to trust what an AI says about health

AI chat assistants can be genuinely useful for making sense of your own patterns. They can also be confidently wrong. The World Health Organization has urged caution in using large language model tools for health, warning that their answers can seem credible while being incorrect or containing serious errors.^[2]

A study published in 2026 had physicians review 888 chatbot answers to 222 questions real patients might ask, including women’s health questions, and found that between 21.6% and 43.2% of answers were problematic depending on the model, and between 5% and 13% were unsafe.^[3] The best-performing model in that study came from Anthropic, the company behind Bloom AI, and it still got a meaningful share wrong. One failure the reviewers described was an assistant assuming a person asking about bleeding was still having periods, without asking whether she had been through menopause.^[3]

Bloom AI is set up not to diagnose, not to suggest medicines or doses, and to tell you plainly when something sounds worth taking to a professional. That lowers the risk. It does not remove it. Treat its answers as a thoughtful friend’s view of your own logs, not as a clinician’s opinion.

What Bloom AI can’t do

• It cannot examine you, order tests, or see anything you haven’t logged.
• It is not medical advice, and it can make mistakes.
• It is not a fertility tool or a method of contraception, and it cannot tell you whether you ovulated.
• It has no way to send your history or a summary to your doctor.
• It won’t remember everything forever: it works from a recent summary of your record, not every entry you have ever made.

Getting useful answers

• Log first, ask second. A few weeks of check-ins give it something real to work from.
• Ask about patterns, not diagnoses. “Is my energy always low before my period?” plays to its strengths. “Do I have PMDD?” is a question for a doctor.
• Leave out what you don’t want sent. Anything you type goes to Anthropic to be answered, so keep identifying details out of your messages.
• Check anything that matters. If an answer would change what you do about your health, confirm it with a doctor, nurse or pharmacist.

Bloom is a wellness and planning app. It is not a medical device, it does not diagnose or treat any condition, and it is not a method of contraception. This page is general information, not medical advice. If something about your cycle worries you, talk to a doctor, nurse or midwife.