They do placebo-controlled trials of therapy techniques like CBT, although for an educated patient it may not be a great placebo - similar to the problem with trying to placebo psychedelics. But basically you can assign the patients randomly to either an informational session on the disorder + unstructured talk "therapy", or the actual therapy method such as CBT.

CBT does perform quite well repeatedly in these sorts of experiments, particularly for depression (it is less studied in some other disorders although still promising). You're right there can be a lot of variance between people though. There can also be additional variance IRL due to therapist quality IMO.