How to read health information without being misled

A polished headline, professional title or confident AI answer does not establish that a health claim is reliable. Safer reading means checking where the claim came from, what evidence supports it and whether it applies to the decision you face.

Start with the exact claim Health information appears in news reports, adverts, podcasts, social media, patient stories, research papers, guidelines and AI answers. These formats do not carry the same purpose or level of scrutiny. A health claim is a statement about health, risk, diagnosis, prevention, treatment or prognosis. Rewrite the claim in plain language before judging it. Ask what action the claim encourages. Advice to buy a product, stop treatment, avoid vaccination or delay care deserves especially careful checking. Misinformation is false, inaccurate or misleading information that may be shared without an intention to deceive. You do not need to prove someone's motive before checking whether a claim is supported. Check the source, but do not stop there Identify the author, organisation and original publisher. Relevant qualifications can help, but a title, uniform, large following or verified account does not make every claim correct. Look for an editorial process, clinical review, named evidence sources, a publication or review date and a way to report corrections. Check whether the information is intended for your country or healthcare system. A credible source is transparent about methods, evidence, uncertainty, funding and mistakes. Credibility raises confidence, but no source should be treated as infallible. Website endings and professional design are only clues. A commercial site can contain accurate information, while an official page can become outdated or may not apply in another country. Identify what kind of evidence you are reading An anecdote is one person's account. It can reveal experiences, priorities or possible harms, but it cannot show how often an outcome occurs or prove what caused it. An advert or testimonial is designed to persuade. It may select favourable facts and leave out uncertainty, alternatives or harms. A preprint is a research manuscript shared before formal peer review. It can provide useful early evidence, but its methods and conclusions may change after further scrutiny. Peer review means other specialists examine a manuscript before publication. It can identify problems, but it does not guarantee that a study is correct, unbiased or clinically important. A single study can be important. Its strength depends on the design, methods, sample, outcomes, missing data and whether the findings have been reproduced. A systematic review uses explicit methods to find, assess and summarise studies answering a defined question. It can provide a broader view, but weak, missing or very different studies can still limit its conclusions. A guideline combines evidence with judgements about benefits, harms, values, feasibility and resources. Check who produced it, the population covered, its date and whether it applies where you live. Match the evidence to the claim Laboratory and animal studies can explain mechanisms and guide later research. They do not establish that the same benefit or harm occurs in people. Observational studies can identify an association between an exposure and an outcome. They are essential for many questions, especially when randomised trials would be impossible or unethical. An association does not by itself establish causation. Confounding occurs when another factor is linked to both the exposure and the outcome and partly explains the observed relationship. Reverse causation can also occur. The outcome or an early stage of illness may influence the supposed cause rather than the other way round. A well conducted randomised trial can provide strong evidence about the effects of an intervention. It can still be limited by bias, short follow up, missing outcomes or a narrowly selected population. No single study design is best for every question. The appropriate design depends on whether the claim concerns treatment, diagnosis, prognosis, prevalence, harms or causes. Read beyond the headline Headlines and short videos often compress a complex result into one sentence. Find the original study, guideline or a reliable evidence summary when the claim could affect your health. Check whether the headline describes the pre specified main outcome or a secondary outcome, subgroup or laboratory measure. A statistically noticeable change may still be too small to matter clinically. Look for the comparator. Improvement after an intervention does not show what would have happened with standard care, no added treatment or time alone. Check how many people started the study, how many completed follow up and whether harms were reported. Missing results and selective reporting can change the apparent conclusion. Put risk numbers back into context A relative risk describes a proportional comparison. An absolute risk describes the chance of the outcome in a group over a stated period. A claim that risk doubled is incomplete without the baseline risk. An increase from 1 in 10,000 to 2 in 10,000 is a doubling and one extra outcome in 10,000. Compare risks using the same denominator and time frame. Also check the confidence interval, which shows the statistical precision around an estimate. Statistical significance does not show that an effect is large, clinically important or free from bias. Benefits and harms need the same careful presentation. Check relevance and currency Applicability means how well the evidence fits the person, population, dose, setting and decision in question. A result from one group may not transfer unchanged to another. Check age, health conditions, pregnancy status, treatment dose, follow up period and healthcare setting when these details matter. Average results do not predict one person's outcome with certainty. Evidence changes. Look for the publication date, latest review date, corrections, updated versions and any retraction. A retraction is a formal notice that a publication should no longer be treated as part of the reliable research record. Retraction can follow error, misconduct or another serious problem. Treat commercial claims with proportionate caution A conflict of interest exists when financial, professional or personal interests could influence judgement. Disclosure allows readers to assess that risk. A conflict does not automatically make a claim false. Its absence does not make weak evidence reliable. Methods, complete reporting and independent confirmation still matter. Be cautious with urgent sales pressure, affiliate links, hidden sponsorship, guaranteed results and claims of a secret cure. Phrases such as clinically tested, detoxes or boosts immunity are too vague without a defined outcome and suitable evidence. Natural does not mean safe. A natural product can cause side effects, interact with medicines or contain an unreliable dose. UK rules restrict misleading medicine advertising and unauthorised health claims for foods and supplements. Regulatory compliance does not prove that a product is suitable for a particular person. Do not buy prescription medicines from social media, messaging apps or unknown websites. Use a registered pharmacy and seek appropriate professional assessment. Check AI outputs and digital media Generative AI can explain language and help organise questions. It can also produce hallucinations, which are convincing but false or unsupported statements. AI can invent references, confuse dates and combine reliable and unreliable material. Open and verify each important source rather than trusting a citation because it looks academic. An AI answer is not automatically current, clinically validated or tailored to the person. It should not be used to rule out serious illness or replace urgent assessment. Images, charts and videos can also be edited, cropped or removed from their original context. Check the original publisher and full material before drawing a conclusion. Use a practical ten question check Before acting on or sharing a health claim, check several things. Consider what exactly is being claimed and what action it encourages, and who produced it and what relevant expertise and editorial checks are visible. Note when it was published or reviewed and which country or population it covers, and whether it is an anecdote, advert, preprint, study, systematic review or guideline. Try to find the original evidence rather than only a headline or screenshot, and consider whether the study design supports the conclusion being claimed. Look at the absolute numbers, time frame and uncertainty, and whether benefits, harms, alternatives and missing information are presented fairly. Finally, check whether funding, sponsorship or other conflicts of interest are disclosed, and whether current independent sources reach a similar conclusion or explain why they differ. Act safely when information could change care Do not stop a prescribed medicine, begin an unregulated product or postpone needed care solely because of a post, testimonial or AI answer. Bring the exact link, image or product name to a pharmacist or relevant healthcare professional. They can assess the claim against your medicines, health background and current guidance. If someone may be in immediate danger, use the appropriate emergency service rather than continuing to search online. Good health literacy supports safer questions and decisions. It does not turn general information into a personal diagnosis.

Judge a health claim through several checks. Identify the exact claim, source, evidence type, numbers, uncertainty, relevance, date and interests before acting or sharing.