(Original Article: https://brucerlong.blogspot.com/2025/05/therapists-are-not-ai-proof-why-because.html)
Does an AI therapist or counsellor need the same set of soft skills and rapport building capabilities/practices as a human therapist?
It's very unlikely.
In fact: no. It doesn't.
But it can deploy near flawless soft skills anyway.
The rapport building is not necessary because human trust is not needed in such a situation. The GPR is just reliable and non-judgmental.
Therapists need rapport building because the client knows they're dealing with a human with human emotional and moral habits and attitudes, and with all of the properties which make negotiating asymmetric interpersonal relationships difficult. Judgementalism. Moral umbrage. Superior intellect. Condescension. Intrinsic bias. Patronising views. Politics. Gender bias. The list goes on.
Human clients - including young people - know that most such dynamics and issues do not matter with an AI GPR, but that the AI GPR can still provide enormous epistemic-therapeutic resources and emotional support, including emulating empathy if necessary.
The client doesn't have to worry about what the therapist really thinks about them, but can still rely upon it having better knowledge, skills, and engagement than most error-prone and flawed therapists.
Forbes overestimates how AI proof teaching roles and therapy roles are...
Philosophy and perhaps performance art might be an exception because of the abstract and complex nature of many of the concepts involved. But all other teaching roles which involve teaching mostly quantitative and procedural skills (law, finance, most sciences) are vulnerable to replacement with an AI GPR.
I also think that the therapy role optimism is misplaced. There are several reasons for this:
1. Iatrogenic injury
Iatrogenic injury is caused not only by physicians by but psychiatrists and psychologists, and the rate of II is very high. AI and AI GPRs are far less likely to make the kinds of errors that lead to iatrogenic injury. That reason alone is enough.
2. Therapist Quality
A lot of human therapists might not cause iatrogenic injury, but are just not very good, and figuring this out can take a long time and cost money and time. Quality control is much easier with AI agents and AI agentive GPRs.
3. Therapist Fit is Solved
It is not easy to find a human therapist who is the right fit. This IS something that can be solved by AI which a) has massive resources for assessing what the client needs and b) can use almost perfect knowledge of all salient theories and methodologies in care delivery to adjust its approach and response, constantly and near-flawlessly measuring behavioural and emotional feedback from the client.
The AI therapist can change to fit the client's needs easily and dynamically.
4. Rapport building is NOT NEEDED
Rapport building and human trust is unnecessary for a human client of an AI GPR because human clients - including children - know that they don't need to worry about the social and interpersonal safety and comfort of the therapist. Assuming otherwise is a mistake. The human client knows that they will get all of the answers and guidance required on the basis of far more complete knowledge than a human therapist can offer without any need to worry about human awkwardness, judgement, moral disgust, or any kind of ick factor.
The AI GPR therapist doesn't have to convince the client that it's their friend and likes them, because the client doesn't need to care about it when they're not being assessed by a human. This doesn't necessarily negatively affect the quality of therapy, and to think otherwise is almost certainly a mistake.
The need for human rapport is just an interpersonal roadblock or bottleneck that is removed with an AI therapist. Nonetheless, a good AI GPR will have the ability to induce calm by way of very good tone, emotional responses, and gently spoken assurance. The massive available expertise levels alone are comforting. The client will be aware that the AI GPR has vastly more resources to draw upon and can do so quickly and accurately. Integration with facial recognition systems that read facial cues by using face image deep learning systems and similar systems for body language already exist.
