The issue here is what is meant by “works.” The placebo effect is a term that encompasses a wide variety of improvements in symptoms measured during a research study that cannot be due to actual physiological effects of the intervention. Some of these effects are real, such as the phenomenon known as regression to the mean, in which symptoms really do get better but only as part of the natural course of a chronic disease with waxing and waning symptoms. Whatever one is doing when the symptoms improve gets the credit for the improvement, even though the change is really just spontaneous. Another example is that subjects in a clinical trial tend to do a lot of things they don’t normally do besides take the intervention being studied: they take their other medications more faithfully, follow diet and other lifestyle guidelines, and so on. This is the Hawthorne effect, and it often makes experimental interventions, including placebos, look like they are doing something.
Other so-called placebo effects are not real improvements but artifacts of data collection: random chance, a tendancy for patients to tell doctors and researchers what they want to hear, and the tendancy of researcher bias to influence the nature of subject assessments.
In any case, clinical trials tend to get positive results for a lot of reasons that have nothing to do with the actual action of the intervention. The mistake is to think that because subjects in a trial appeared to improve while taking a known placebo that somehow the act of taking the pill, even knowing it was a fake, led the mind to somehow heal the body. Far more likely is that a collection of factors such as those I described above, and many others I didn’t go into, led to an apparent or real improvement in symptoms that had nothing to do with any actual lasting change in the disease. It is clear from other research that such apparent effects 1) are usually short lasting, certainly much shorter than those seen with actually effective treatments, 2) are small and often not clinically meaningful, and again nothing compared to the effect sizes for real therapies, 3) are usually not seen with acute, severe, and objectively measurable problems; you won’t find placebos useful for preventing pregnancy, curing sepsis, or treating the pain of a broken leg, but you will find apparent results when used in chornic diseases with mostly subjective symptoms.
I don’t doubt that people sometimes feel better because of interventions, rituals, and lots of other things that don’t actually affect theri disease, and there’s nothing wrong with patients seeking such remedies, The problem is with healthcare providers fooling themselves into thinking there is some “mind-over-matter” magic involved that has real healing properties, which just leads them to offering people useless therapies and which sucks energy, talent, and reseources away from the effort to find truly effective treatments.