In short
An observational study is a research design in which researchers measure and analyze outcomes as they naturally occur, without assigning participants to a particular exposure or treatment the way a randomized controlled trial does.
Common types include cohort studies (following a group over time and comparing outcomes between those with and without a given exposure, such as the long-running Framingham Heart Study or Nurses' Health Study), case-control studies (comparing people who already have an outcome of interest to those who don't, looking backward for differing exposures), and cross-sectional studies (measuring exposure and outcome at a single point in time). Because researchers don't assign who gets which exposure, people naturally sort themselves into groups based on their own choices and circumstances.
This self-sorting is the central limitation: people who choose a given behavior, supplement, or lifestyle factor often differ systematically from those who don't, in ways that are difficult to fully measure and account for statistically (see: confounding variable). A frequently cited illustration is the "healthy user bias" seen in some observational nutrition and supplement research, where people who take a given supplement also tend to exercise more, eat better, and see doctors more regularly — so an association between the supplement and better health outcomes may partly or entirely reflect these other correlated behaviors rather than the supplement itself.
Observational studies remain genuinely valuable — they can study outcomes over far longer timeframes and larger populations than an RCT practically can, and are often the only ethical way to study a suspected harm — but they are used in this site's evidence framework primarily to generate hypotheses and identify associations worth testing, not as standalone proof that a specific intervention causes a specific benefit, which is why they place a compound in a lower evidence grade than repeated positive RCT results.