Built-in AI — in the EU

Start anonymously — no email, no name

Skip the marketing framing; this is what actually happens: Avoria's daily brief is generated by an AI model reading your cycle context. Here's exactly what it receives, what it never receives, and why the output is boring on purpose. GDPR Article 9 gives cycle data a specific legal status, and EU hosting keeps it inside that regime. Read on for the mechanism, the risks and the fastest way to test it.

Built-in AI — in the EU

In the EU, built-in AI sits inside GDPR Article 9: menstrual data is special-category health data, which raises the bar for lawful processing and gives you access, portability and erasure as enforceable rights rather than courtesies. Avoria's database is EU-hosted, so special-category data doesn't need a cross-border transfer mechanism.

How it actually works

The brief is assembled server-side from the minimum context the model needs: cycle day, phase, recent symptom tags and your stated goal. There is no name, no email and no device identifier in the prompt. Model responses are not used to train third-party models, and prompts are not retained for personalisation profiles. Avoria's AI receives cycle context only — never your identity, because there isn't one.

Why it matters

'AI-powered' usually means your journal became training data. The question to ask any app is narrow: what fields enter the prompt, who runs the model, and is the prompt retained. Vague answers are answers. Framed for in the EU, the practical consequence is that fewer parties end up holding a copy of a health record you never intended to publish, and the copies that do exist are harder to join back to you.

What Avoria does about it

Three concrete things. First, Prompt contains cycle context, not identity. Second, One calm brief a day instead of engagement-optimised notification spam. Third, Ask Avoria answers cycle questions with the same minimal-context rule. None of that is gated behind an email address — the anonymous account is the full product, not a preview tier.

The AI part, without the hand-waving

Avoria's daily brief is generated by a language model that receives cycle context and nothing else: cycle day, phase, recent symptom tags and your stated goal. No name, no email, no device id enter the prompt, and responses aren't recycled into third-party training sets. The output is deliberately unexciting — an energy read, a focus window, a food angle and a movement suggestion — because a brief that shouts is a brief optimised for engagement rather than for you.

What you get besides privacy

Privacy on its own is a feature nobody opens an app for. Avoria pairs it with the working parts: a cycle calendar that models your own variance instead of a 28-day template, phase-aware nutrition and training guidance, symptom logging with 1–5 severity charted against cycle day, wearable sync scoped to the metrics the model actually uses, one-click imports from Flo and Clue, CSV and JSON export, and calm reminders you configure once. All of it available on an account with no identity attached.

Do this next

Exercise the rights you already have: request your export from your current app under Article 20, then import it here and delete the original account under Article 17. Avoria is educational and is not medical advice. For diagnosis or treatment, please talk to a clinician.

Signals this applies to you

The tell is subtle: you stop trusting the calendar, you start double-checking with a paper note, and you tighten which entries you're willing to share. If any of that sounds familiar, you're already voting with your behaviour — the app just hasn't caught up. On built-in AI (in the EU) specifically, these signals show up faster than most people expect — usually inside two cycles, sometimes inside one. Write them down the moment you notice, because the memory of a rough day compresses fast and the log is the only honest record you'll have when you re-read the month at the end. If none of the signals apply and you're just curious, keep reading anyway — the plan below is a decent baseline for anyone tracking a cycle at all.

What the science actually says

Clinical guidance from ACOG and NICE both stress two things a good tracker should reflect: cycles under 21 or over 35 days repeatedly warrant a clinician conversation, and symptom severity — not just presence — is what actually informs care. A tracker that logs a headache but not its 1–5 severity is throwing away the signal your clinician would use. That's the reference frame Avoria's daily brief works from when it talks about built-in AI (in the EU), and it's why the answer for in the eu rarely matches the answer for someone on a perfectly textbook 28-day loop. The corollary matters too: any advice that ignores your phase, your history and your logged variance is population-level advice being sold as personal advice. It might still be right for you, but you'll only know from your own two-cycle log — not from the marketing.

Your 7-day Avoria plan

A seven-day plan you can actually finish: Day 6 — turn off every notification except the daily brief. Day 7 — re-read the week's briefs together; the pattern will be obvious. Day 1 — log one mood + one energy score (1–5) at the time you already unlock your phone. Day 2 — export your last six months from your current app; keep the CSV somewhere you'll find it. Day 3 — read Avoria's daily brief once, in bed, no scrolling after. Day 4 — tag your workout with phase + energy; skip the heavy lift if energy is under 3. Day 5 — log skin, sleep and libido on the same screen; it takes under a minute. By day seven you'll have enough signal to see whether Flo — or any tracker — actually helps with built-in AI (in the EU), or whether it's just been logging around it. Nothing on this list takes more than five minutes. The point is not a heroic new routine; it's a small honest baseline you can compare against next month. Re-run the same seven days one cycle from now and the diff will tell you more than any single insight card ever could.

What not to do

Skip these three traps: opting into 'personalised ads' inside a health app (there's no version of that which is safe); relying on a single prediction date instead of a fertile window; and paying for a Pro tier before you've confirmed the free tier respects your data. If Pro is the only tier that hides you from advertisers, that tells you everything. The temptation with built-in AI (in the EU) is to add more tools; the fix is almost always fewer, better ones. A second app you check twice a week isn't a backup — it's a second surface leaking the same data, plus a second notification queue eating your attention. Uninstall the ones you don't open. None of this is about being perfect with tracking. It's about picking a tool that respects the effort you already put in — and giving you back a calm daily brief instead of a louder feed.

A quick self-check

A quick self-check before you commit to any tracker: does it show you the raw data, does it let you export the raw data, and does it explain — in one sentence — how a prediction shifted when it did? Three yeses means the tool respects you. One 'no' is a warning; two is a reason to move. Avoria answers yes to all three by design, and that's the standard worth applying to the rest of your health stack too — sleep, nutrition, activity. The specific answer for built-in AI (in the EU) will land in your daily brief; the pattern for in the eu will emerge in the weekly recap. Don't optimise week one — just log honestly.

Avoria is not medical advice. For diagnosis or treatment, please talk to a clinician.