Sorry, but I wonder if there is a fundamental bias in the investigation: you wish to compare human therapist with voicebots, however you seem to wish to do your research online?

Given that anything online can be hacked, copied, stored, it would seem to me that this shows already that human therapist settings are more confidential -

I am not sure when you say that you want to compare human therapists with voicebots, if confidentiality should be the biggest concern. In 2020, when most of the world went into lockdown, and most therapy was online, suicide went up by over 20% worldwide - which might on top of peoples fears and anxieties and depression and isolation maybe suggest, that the lack of human contact had a major role to play in suicide -

In a real client/therapist relationship Confidentiality or "Privacy" is kept between the therapist and client unless the therapist is led to believe A.) the client is at immediate risk of harming themselves or another individual, B.) there is abuse of neglect of a child or C.) there is abuse or neglect of a vulnerable adult (ex. elderly, or disabled individual.) Anything else for any other reason is kept strictly between the therapist and client. When the therapist does confide in another agency to ensure client safety, only the imperative information is given and anything else is kept private.

The only other reason client confidentiality would be broken is if the court-mandated a release of therapist files about a client.

Overall, I don't think there is a scale you can use in terms of privacy/confidentialty because breaking confidentiality is never an easy decision for a therapist and requires context in every scenario. I'm not sure there would be the same consideration when comparing to a bot. There is value in human connection and in the education therapists commit themselves to for so long.

On this matter, it would be great if you could create your own scale based on empirical data obtained from the target population.