Period tracking apps hold some of the most sensitive data a person can generate, and AI makes that data more revealing. From cycle logs, an algorithm can infer likely pregnancy, pregnancy loss, fertility treatment or menopause, even when a user never enters those words. In most countries, including the United States, legal protection for that data is patchy.
Imagine a user who has logged her period every month for three years. In March she logs nothing. In April she logs nothing. In May she opens the app, logs bleeding and deletes a symptom note. She has written nothing about pregnancy. She does not need to. Any system that sees the sequence can draw a conclusion about what probably happened, and so can anyone that system shares data with.
Period and fertility tracking apps are used by tens of millions of people. They are genuinely useful: they help people understand their cycles, plan or avoid pregnancy, spot irregularities worth mentioning to a doctor, and feel some control over their bodies. The privacy debate around them often focuses on what users type in. The argument here is that the larger risk lies in what can be inferred. Machine learning turns routine logs into predictions about the most intimate events of a person’s life, and neither users nor most privacy laws are built around that.
What data do period tracking apps actually collect?
The obvious fields are cycle dates, flow and symptoms. Many apps also ask about sexual activity, contraception, ovulation test results, basal body temperature, mood, weight, medications and pregnancy intentions. Some connect to wearables for sleep, heart rate and temperature. Then there is the metadata every app generates: device identifiers, location, usage times and patterns of engagement.
Several of these data types are sensitive in themselves. Combined over time, they become something else: a detailed physiological and behavioural record from which pregnancy, miscarriage, abortion, fertility treatment, sexual activity, hormonal changes and menopause can often be estimated.
That is not a hypothetical capability. Cycle prediction is the core function of these apps; spotting deviations from the predicted pattern is what the algorithms are designed to do. The same modelling that tells a user her next period is due also recognises when it did not come.
The general principle has been public for more than a decade. In 2012, a New York Times Magazine investigation described how a large US retailer had developed a model to estimate from purchase patterns whether a customer was likely pregnant, so it could target marketing. Shopping baskets are a weak signal compared with a daily log of cycles, symptoms and sexual activity.
Who can see period tracking app data?
The short answer is: more parties than most users assume.
The best-known case involves Flo Health, maker of one of the most popular period and ovulation trackers. In 2021, the US Federal Trade Commission finalised a settlement over allegations that, despite promising to keep users’ health data private, Flo had shared sensitive health data from millions of users with marketing and analytics firms, including Facebook and Google. The order required Flo to obtain users’ explicit consent before sharing health information, to notify affected users and to instruct third parties to destroy the data.
A class action followed in California. According to reporting on the case, Google and Flo settled, and in 2025 a jury found Meta liable for collecting users’ reproductive health information through the app.
Flo is not an outlier in its sector. A 2022 study in JMIR mHealth and uHealth by Alfawzan and colleagues analysed 23 popular women’s health apps and found that 20 shared user data with third parties, all 23 allowed behavioural tracking, and only 16 displayed a privacy policy at all.
Third-party software development kits and advertising pixels are the usual channel. An app may never “sell” data in the everyday sense, yet send events such as “user logged period” or “user opened pregnancy mode” to analytics and advertising services. We covered how this happens in our piece on health apps, HIPAA and ad pixels.
Then there is law enforcement. After the US Supreme Court’s Dobbs decision in June 2022 ended the federal constitutional right to abortion, the question of whether reproductive data could be used in prosecutions became urgent. In 2022, police in Nebraska obtained Facebook messages between a mother and her teenage daughter through a warrant in a case involving an abortion. That case did not involve a period app. But it showed that digital records held by technology companies can be obtained through legal process, and that companies will often comply.
Why does AI make reproductive data riskier?
Three features of AI change the picture.
Inference from ordinary data. Traditional privacy rules focus on sensitive data that users provide, such as a diagnosis or a pregnancy test result. AI creates sensitive conclusions from data that looks mundane. A gap in period logging, a change in app usage, a new pattern of symptoms or a wearable showing a sustained rise in resting temperature can together suggest pregnancy. The inference may be wrong, and a wrong inference can also cause harm. Either way, the user never disclosed it.
Scale and linkage. Data flowing to advertising networks and data brokers can be linked with location data, purchase histories and other app data. Location data showing a visit to a clinic, combined with cycle data, is far more revealing than either alone.
Training and retention. Apps adding AI features, from chatbots to predictive insights, may use users’ logs and conversations to train or improve models. Data used for training can be hard to delete fully, and the privacy terms governing it are often vague.
None of this requires bad intent. A company can build excellent predictions to help users and, in doing so, create a sensitive data asset that is valuable to advertisers, attractive to hackers and reachable by subpoena.
A worked example: one user, three data flows
Hypothetical, for illustration. A user in a US state with strict abortion laws uses a free period tracker. She has cycle logging and symptom tracking on, location permission granted for “local weather”, and a connected wearable.
In flow one, the app’s algorithm notices a missed period and a rise in resting temperature and quietly switches her cycle prediction to “possible pregnancy”. She never sees this label, but it sits in the company’s database.
In flow two, the app’s analytics software development kit reports in-app events to a third-party analytics provider, including that she opened the pregnancy content section. Combined with her advertising identifier, this reaches ad networks, and she starts seeing baby product ads.
In flow three, six weeks later, she logs heavy bleeding. The algorithm’s pregnancy label disappears. Her location history, held separately by a data broker that bought it from another app, shows a trip to a city with an abortion clinic.
She typed nothing about pregnancy or abortion. Every piece of data was collected for a plausible reason. The inference, however, now exists in several places, and in some legal environments it could matter a great deal.
What protects period tracking data?
In the United States, the answer is a patchwork. HIPAA, the federal health privacy law, generally applies to healthcare providers, insurers and their business associates, not to consumer apps a user downloads directly. The FTC can act against unfair or deceptive practices, as in the Flo case, and against unauthorised disclosure of health data under its Health Breach Notification Rule. Some states have enacted consumer health data laws; Washington’s My Health My Data Act, for example, restricts collecting and sharing consumer health data without consent. Protection varies widely depending on where a user lives.
In the European Union and the United Kingdom, the General Data Protection Regulation treats health data and data concerning a person’s sex life as special category data, requiring stronger justification for processing. That offers more protection, but enforcement depends on regulators’ resources, and inferred data raises unresolved questions.
The strongest counterargument to this piece is that these risks are overstated. Law enforcement has rarely, if ever, been publicly shown to rely on period app data in an abortion prosecution. Most apps have tightened their practices since 2022, several offer anonymous modes and local storage, and the benefits of cycle tracking are real. Scaring people away from useful health tools has its own cost.
That argument is partly right, and the benefits are real. But the absence of a known prosecution does not measure risk; data practices change, companies are acquired, laws change and breaches happen. Once an inference exists on a server, the user has no control over its future. The better answer is not to abandon tracking but to design apps so that sensitive inferences are never created or stored off the device in the first place. We discussed why ownership framing misses this point in our piece on who controls patient data.
What this means for users, companies and policymakers
For users, check whether an app stores data locally or in the cloud, whether it offers an anonymous mode, which third parties it shares data with, and whether it uses your data to train AI. Turn off location permissions the app does not need. If privacy is a serious concern for you, consider apps that keep data on the device, and remember that your clinician can help with cycle concerns directly.
For companies, make on-device processing the default, avoid creating pregnancy or pregnancy-loss inferences on servers, remove advertising and analytics kits from health features, and publish transparency reports on law enforcement requests. Do not train models on user logs or chats without clear, opt-in consent.
For policymakers, extend health privacy protections to consumer health data wherever it is held, cover inferences explicitly, restrict sale and sharing of reproductive data, and set clear limits on how such data can be obtained through legal process.
For clinicians and health systems recommending apps, prefer those with strong privacy practices and explain the trade-offs to patients.
Frequently asked questions
Is period tracking app data private?
Not always. In the US, most consumer period apps are not covered by HIPAA. A 2022 study of 23 popular women’s health apps found 20 shared data with third parties. The FTC settled with Flo Health in 2021 over allegations it shared health data with firms including Facebook and Google. Check an app’s privacy policy, storage and sharing settings.
Can AI tell if I am pregnant from my period app?
Algorithms that predict cycles can also detect when a period does not arrive, and combined with symptoms or wearable data, can suggest a likely pregnancy or pregnancy loss, even if you never record it. Whether the app creates, stores or shares such inferences depends on its design, which is why on-device processing matters.
Can police get period tracking app data?
Law enforcement can seek data from technology companies through warrants and subpoenas. There is little public evidence of period app data being used in abortion prosecutions so far, but other digital data, such as private messages, has been used. Apps that store data only on your device reduce what can be handed over.
How can I use a period tracker more safely?
Choose an app that stores data on your device or offers an anonymous mode, limit permissions such as location, avoid linking it to social media accounts, and review whether it shares data or uses it for AI training. Periodically delete data you no longer need. For health concerns about your cycle, speak with a clinician.
Period tracking apps help millions of people understand their bodies. The privacy risk is no longer only about what users type in. It is about what machines conclude from it, where those conclusions are stored and who can ask for them.
Do you use a period or fertility tracking app, and has anything about AI or data sharing changed how you use it?
