Good ones are, within limits. Well-built trait scales such as the Big Five give similar scores when the same person takes them weeks apart, agree with what friends say about that person, and predict real outcomes such as job performance, relationship stability and longevity with small to moderate effects. Their limit is that they measure how you see and present yourself, which you can shade or fake.
Reliability: do you get the same answer twice?
Reliability has two common measures. Internal consistency (coefficient alpha) asks whether the items of a scale hang together; test-retest reliability asks whether the score holds over time. The 120-item Big Five test on this site has facet alphas between .63 and .88 from a sample of 619,150, and the 50-item markers have scale alphas between .79 and .87. Test-retest correlations for Big Five scales are typically in the .7 to .9 range over weeks and months. Shorter scales are less reliable: the ten-item quick test has alphas as low as .40, as its authors expected, and is sold on its test-retest figures instead (Gosling, Rentfrow and Swann, 2003). Every test page here states its figures and the methodology page lists them all.
Validity: does it predict anything?
Roberts and colleagues (2007) reviewed longitudinal studies and found that personality traits predict mortality, divorce and occupational attainment about as well as socioeconomic status and cognitive ability do. Soto (2019) then tried to replicate 78 published links between traits and life outcomes in a pre-registered study and found that most held up, with effects typically smaller than first reported. So the traits are real and consequential, and the effects are modest: a trait tilts the odds across a population, it does not decide anyone’s life.
Faking, and what self-report cannot see
Any self-report can be faked, and people asked to answer as a job applicant would can raise their scores by a standard deviation or so. That is why this site will never help an employer screen candidates, and why the comparison samples are people who answered for themselves. Self-report also reflects self-image, mood on the day and how a person wants to be seen, which is both a limit and part of what is being measured.
What it means for your results
Trust a Big Five result as a fair reading of how you see yourself, precise to within a few points and a few percentile ranks. Trust a two-item scale less. Trust any test less when you answered as you would like to be, and not at all when it sorts you into a type.