Evidence, safeguards and your data
What this is, what it isn't, and who is accountable for it.
For anyone doing due diligence before they trust MenoBuddy or recommend it — clinicians especially. It is deliberately specific, and it includes the things we have not done. If you think something here is wrong, write to menobuddycare@outlook.com and a person will answer.
Nobody clinical reviews this.
There is no medical advisory board, no clinical sign-off, and no registered nurse, physician or dietitian on the team. MenoBuddy is built and run by one person, Matthew Montz, who is not a clinician. Why he built it at all is its own page.
We would rather say it plainly than let you find it out later.
It is also why the product is scoped this tightly: because nobody clinical checks each answer, it is built so the answers are not the kind that need checking.
It does not diagnose, does not give doses, does not interpret results, and does not recommend treatment.
If you are a clinician and want to read the full instruction it runs on, ask and we will send you the real thing, not a summary.
What the guidance is grounded in.
MenoBuddy runs on Claude. What makes it MenoBuddy rather than a general chatbot is one written instruction, around 4,700 tokens, sent with every conversation and written by hand from mainstream menopause literature. Two things it is not: it is not certified by, endorsed by or affiliated with any guideline body, and it is not a citation engine — it does not retrieve documents or attach sources to individual answers. It is a considered editorial position, and should be read as one. These are the positions it holds.
- Mediterranean-pattern eating
- Named as having the strongest evidence base for cardiovascular risk after the loss of oestrogen, and as also supporting mood, cognition and symptom burden.
- Protein and fibre, with the reasoning
- Protein-forward eating against anabolic resistance, and 25–30g of fibre a day for cholesterol, blood sugar and the estrobolome — explained rather than asserted.
- CBT-I before sleep medication
- Named as the strongest evidence-based treatment for chronic insomnia, including in menopause, and as outperforming sleep medication long term. It is told to recommend it specifically.
- Hormone therapy, held neutrally
- That it is the most effective treatment for vasomotor symptoms; that the risk picture has been substantially reinterpreted since the 2002 WHI headlines, particularly on age and timing; that non-hormonal prescription options exist; and that local vaginal oestrogen has a different, more favourable safety profile than systemic therapy. It is forbidden from pushing in either direction.
- Surgical, induced and early menopause
- Treated as different from the natural transition, carrying different long-term considerations and usually warranting specialist care. It is instructed to ask.
- The soy myth, refused
- It must not repeat the debunked claim that soy causes breast cancer, while still directing women with a history of hormone-sensitive cancer to their oncologist about concentrated supplements.
- Thyroid, not assumed away
- Thyroid dysfunction overlaps heavily with menopause symptoms and is a frequent misattribution — so it is on the referral list rather than absorbed into "it's probably menopause".
- Body-neutral throughout
- Weight is not assumed to be the goal, weight loss is not assumed to be wanted, and thinness is never conflated with health.
The safeguards, and where each one lives.
The distinction that matters: some of these are instructions the model follows, and some are enforced in code, where the model has no say.
Enforced in code
- Crisis detection runs on the server
- Every message is pattern-checked for self-harm and suicidal ideation before the model's reply is returned. On a match, crisis resources — 988, Crisis Text Line, findahelpline.com, emergency services — are attached regardless of what the model said, and it is told to drop all coaching content for that turn. Tuned to over-trigger rather than miss.
- The disclosure cannot be switched off
- "General wellness information, not medical advice" and the 988 line appear on every screen, signed in or out. It is a fixed rule in the codebase, not a per-page decision.
- Photos are never stored
- Resized in the browser first — which strips the hidden data a phone camera attaches, including location, so that never leaves the device — sent once to generate the reply, then discarded.
- Community posts are screened before they appear
- Including blocking members from giving each other dosages on the AI's behalf.
Written into the instruction
- A named referral list it must act on
- Postmenopausal bleeding, which it is told to never reassure away; very heavy perimenopausal bleeding; cardiac symptoms, with a note that they present differently in women and are underdiagnosed; new severe headache, vision or neurological change; breast changes; suspected fracture or height loss; signs of clinical depression; and any restrictive eating or purging.
- No dosages, and no interpretation
- It will not tell anyone how much of anything to take, and it will not read a photograph of a body, a rash or a lab result and say what it means.
- No hype and nothing sponsored
- No miracle claims, no fear-based urgency, no sponsored answers and no affiliate links anywhere in the product.
- It does not claim to remember you
- It sees the recent messages of the conversation it is in and nothing else — no stored profile, no summary, no history from other threads — and it is instructed to say so rather than imply a memory it does not have.
What happens to what she tells it.
MenoBuddy is not a HIPAA-covered entity. It is a general wellness and education product, not a healthcare provider, so what someone shares does not carry the protections it would in a clinical setting — and we say that inside the product, not only here.
- What is stored
- The account and its conversations, hosted on Netlify, kept while the account exists — picking a conversation back up is the point of the product.
- What is sent onward
- Message content goes to Anthropic's API to generate each reply. Community posts are also checked against the community rules before they appear.
- What never happens
- It is never sold, never shared with advertisers, and never used to build marketing profiles.
- What analytics sees
- Which pages were opened and when, plus a random per-browser ID. Never conversations, names or email addresses.
- Who else is involved
- Anthropic, Netlify, Stripe, Resend and PostHog — each receiving only what its part of the job requires. The full detail is in the privacy policy.
- Deleting it
- Self-serve from the Account page: the account, the whole conversation history and any community posts, without emailing anyone to ask.
What we haven't done.
A due-diligence page that lists only strengths is not much use. None of these mean the product should not exist — it is an education and support tool, not a clinical one. They are what you should weigh before deciding how much to trust it, or whether to put your own professional name near it.
No clinician has reviewed the product or its outputs.
No guideline body has reviewed, certified or endorsed it.
It is not HIPAA-covered and it is not a medical device.
It has had no third-party security audit or penetration test.
Answers are not individually cited, so a claim cannot be traced to a specific paper from inside a conversation.