Our Methodology
We believe mothers deserve the truth. That's why we say openly how our content is made, where AI has limits and where you should not rely on us. No marketing gloss.
How our content is created
Every page on ask mamis — whether a guide, a pregnancy week, a baby month, a symptom or a forum answer — starts as a structured prompt sent to a large language model. For editorial content we use Anthropic's Claude Opus; for translations and shorter additions we use Claude Haiku. Each prompt carries three fixed blocks: the mother's concrete question, a list of permitted sources (WHO, AAP, NHS, BZgA, DGKJ, RKI, Cochrane, UpToDate, DGGG, ACOG) and a tone profile that puts empathy before facts. The model is not allowed to invent content outside that source list and must tie every concrete recommendation to a guideline. Before publication, each text is run through a second, independent model that checks three things: does any statement contradict a current guideline, is a mandatory disclaimer missing (fever, vaccination, emergencies), and is the tone bearable for an exhausted mother. Content is generated per language on purpose: an English text is not translated from the German one, it is written from the same source list so that cultural differences (e.g. Hebamme in Germany vs. midwife in the UK vs. doula in Spain) are captured correctly. Numbers, age ranges and dosages are cross-checked against the original source. Despite these steps, a model can still get things wrong — that is why every page carries a visible notice and an email address for error reports.
Our sources
Every statement on ask mamis must be derivable from one of these publicly available sources. We have no advertising ties with these organisations — they don't even know we exist.
WHO (World Health Organization)
Open sourceGlobal guidelines on pregnancy, birth, breastfeeding (exclusive breastfeeding 0–6 months), immunisations and infant health.
AAP (American Academy of Pediatrics)
Open sourceU.S. specialty society for paediatrics. Basis for sleep recommendations, introduction of solids, immunisation schedules and SIDS prevention.
NHS (UK National Health Service)
Open sourceUK public health information for pregnant women and parents — practical, well-structured and freely accessible.
BZgA (Bundeszentrale für gesundheitliche Aufklärung)
Open sourceGerman federal authority — official guides for pregnancy, birth and child health ("Kindergesundheit-info.de", "familienplanung.de").
DGKJ (Deutsche Gesellschaft für Kinder- und Jugendmedizin)
Open sourceGerman paediatric specialty society. Provides guidelines for complementary feeding, routine paediatric check-ups (U1–U9) and prevention of common childhood illnesses.
RKI (Robert Koch-Institut)
Open sourceGerman federal institute for infectious diseases. STIKO vaccination recommendations, data on measles, RSV, pertussis and infection prevention.
Cochrane Library
Open sourceSystematic reviews as the gold standard of evidence-based medicine. Our prompts request Cochrane evidence whenever a topic is contested.
UpToDate
Open sourceClinical decision-support system used by over two million clinicians worldwide. Reference for differential diagnoses in paediatric symptoms.
DGGG (Deutsche Gesellschaft für Gynäkologie und Geburtshilfe)
Open sourceGerman specialty society for gynaecology and obstetrics. Publishes AWMF S3 guidelines for pregnancy, birth and postpartum care.
ACOG (American College of Obstetricians and Gynecologists)
Open sourceU.S. professional body for obstetrics and gynaecology. Source for antenatal care, screenings and risk assessment.
WHO Child Growth Standards
Open sourceOfficial WHO growth charts (0–5 years). Data basis for our growth calculator and reference for questions about weight development.
Stiftung Warentest / Öko-Test
Open sourceIndependent German consumer testing bodies. Reference for product comparisons (car seats, baby food jars, baby monitors, skincare).
What our AI CANNOT do
An AI is not a doctor. It cannot see your baby, feel its forehead or know its medical history. It cannot make an individual diagnosis. It cannot recognise an emergency — if you see blue lips, absence of breathing, severe accidents or seizures, call 911, 112 or your local emergency number immediately, do not check us first. It may not prescribe medication and cannot confirm dosages beyond the general package insert. It does not replace midwife care, a paediatrician appointment, routine check-ups or postnatal follow-up. It is not a therapist: if you suspect postpartum depression, an anxiety disorder or have suicidal thoughts, call your local emergency line or crisis hotline. And it is not live: guidelines change, the model's training snapshot is frozen at a moment in time, and we cannot guarantee that every page always reflects the latest recommendation.
How we prevent errors
We work with six safety nets. First, source-first prompting: before the model answers, we place the relevant guideline into its context — if it invents beyond that, we abort. Second, the mandatory disclaimer rule: for fever, jaundice, breathing difficulty, seizures, bleeding, suicidal thoughts and postpartum depression, an emergency notice is prepended automatically and cannot be disabled. Third, a tone safety net: we measure shame, blame and ableism signals in the text and automatically rewrite sections once a threshold is crossed. Fourth, citation enforcement: for every concrete number (sleep duration, age for starting solids, growth percentile, vaccination interval) the model must name the cited guideline in a structured field. Fifth, localised disclaimers: in Germany we point to the Hebamme and Kinderarzt, in the UK to the health visitor, in Spain to the matrona — each locale references the care structure actually available in that country. Sixth, periodic audits: each month we sample fifty pages at random and check them against current guidelines; deviations are corrected and logged in a changelog. None of these nets makes us error-free. They make errors rarer and easier to surface.
Roadmap: real reviewers from 2027
The most honest sentence we can write: right now ask mamis has no medical reviewers. No midwife, paediatrician or gynaecologist checks the texts before they go live. That is a deliberate current state — and it has an expiry date. From spring 2027 we will bring real professionals on board in three waves. First IBCLCs, internationally certified lactation consultants: this is the most under-served domain, and they will start by signing off breastfeeding content (0–6 months). In summer 2027 midwives follow (in DE, AT, CH, FR, ES and UK), approving pregnancy, birth preparation and postpartum content. By the end of 2027 we want at least two paediatricians per language reviewing the health and symptom pages. Every reviewed page will carry a visible review line with the professional's name, qualification, registration number and date, similar to NHS Inform. Until then, please read everything here for what it is: AI-generated knowledge from public guidelines, not the voice of a doctor.
Help us find errors
If you find a contradiction with a guideline, an outdated number, an unsafe dosage or a tone that slips, please tell us. An email to hello@ask-mamis.com with the subject "Content feedback" and the URL of the page is enough. We reply within five working days, correct the page within two weeks at the latest and log every change in a public changelog. Midwives, doctors, IBCLCs and doulas who have the time and inclination to formally review our texts are also welcome at the same address — we pay per review from 2027 onwards.
Email us · hello@ask-mamis.comLegal notice
ask mamis is not a medical institution, not a health consultation within the meaning of § 1 of the German Healers' Act and not a drug information service. The content is for general information only and does not replace a medical examination or advice from your midwife, paediatrician or OB/GYN. We accept no liability for decisions taken on the basis of this content. For health concerns, consult the regular care structure of your country. In an emergency, call the emergency services immediately (112 in the EU, 911 in the USA, 999 in the UK, 155 in Turkey, 123 in Lebanon, 997 in Saudi Arabia).