How we fact-check

Applies to all 67 published fact-checks

KIDS checks claims about children’s health that parents are hearing in the news, on social media and from each other. Every fact-check is written by AI from published research and official health guidance, then checked by a second, independent AI review before it is published. No doctor reviews them. This page explains exactly how it works and where it can go wrong.

1. Choosing claims

Claims come from two places: trending news and online discussion about children’s health, scanned several times a day, and a list of long-standing parenting myths. A claim has to be specific enough to test, such as “sugar makes kids hyperactive”. We skip news events, product promotions and claims about individual people. We don’t cover COVID-19.

2. Gathering the evidence

For each claim we search PubMed, the US National Library of Medicine’s database of medical research, prioritising systematic reviews, meta-analyses, randomised controlled trials and practice guidelines in children, adolescents or pregnancy.

We also read current guidance pages from official health bodies, including the Royal Children’s Hospital Melbourne, Raising Children Network, ASCIA, NHMRC, healthdirect, Red Nose, the Therapeutic Goods Administration, the American Academy of Pediatrics, the CDC, the NIH, the WHO, the NHS, NICE and Cochrane. Australian sources are preferred.

3. Reaching a verdict

Supported: good evidence the claim is true. Mixed: the evidence conflicts, shows a link without proving cause, or depends heavily on age, dose or circumstances. Not supported: good evidence the claim is false, or it has been tested and not borne out.

Each verdict also gets an evidence strength. Strong: consistent systematic reviews or trials, or clear agreement in official guidance. Moderate: reasonable but limited or mostly observational evidence. Limited: sparse or weak evidence.

The writer may only use the evidence it was given. Any number, study, year or organisation in a fact-check must come from a source listed on that page.

4. Checks before anything is published

Every fact-check must cite at least three sources the system actually retrieved, including at least one systematic review, trial, guideline or official guidance page. Citations that don’t match a retrieved source are rejected.

A separate AI auditor then compares every statement, number and the verdict itself against the full evidence. If it finds a problem, the writer gets one chance to fix exactly that, and the fixed version is audited again from scratch.

Claims about vaccines, medicines, safety and pregnancy also need official guidance among their sources. Anything that fails these checks is held back and not published.

5. What a fact-check is not

KIDS is general information, not medical advice. A verdict describes the overall evidence, not what is right for your child. Fact-checks can be wrong or out of date. Always talk to your GP or child health nurse about your child, and call 000 in an emergency.

6. Corrections

Every claim page has a way to report a problem, in the app and by email to nick@0xlaunch.me. Claims reported by several parents are taken out of the feed while we look again. When the evidence changes, we re-check the claim and update the date it was checked.