The Therapy Industry Is Misreading AI

The Therapy Industry Is Misreading AI

And the pattern of that misreading is not new.

People are using AI for self help at scale. Harvard Business Review’s 2025 study of generative AI use found that therapy and companionship is the number one consumer use case. Organizing life is second. Finding purpose is third. Personal support use nearly doubled year over year and now represents roughly a third of all AI usage.

The therapy industry’s response has been to ring alarm bells. Posts from therapists discounting or criticizing AI. The replies flood in: some genuine, some hostile. Edge cases get shaped into a new diagnosis called AI psychosis. AI proponents wave off real risks as friction. Both reactions miss what is happening.

The therapy industry is responding to AI the way incumbent industries always respond when a new tool removes a bottleneck they control. The first move is to call the threat itself the harm, instead of asking what value the shift created. History gives the pattern: scribes and the printing press, musicians and recorded music, film studios and home video, taxi medallions and ride-sharing. In each case the same sequence. A tool removes a bottleneck. Incumbents frame objections in safety, ethics, or cultural harm. Some of the concerns are real. They coexist with economic self-protection. The profession survives only after redefining human value.

The Therapy Industry holds three bottlenecks: access, interpretive authority, and paid presence time. AI weakens all three by making reflective interaction constant and cheap.

There are legitimate concerns about AI in psychological work. Sycophancy, dependency, reality-displacement, skill atrophy, alignment contamination. Users should understand them. Practitioners should name them. But the public conversation rarely separates those concerns from fear of displacement. When the two are blended, credibility erodes, and the serious question gets lost.

The serious question is not whether AI belongs in psychological work. It already does. The question is who owns the tool when it enters, and what the practitioners who still have the craft choose to do in the time they have to shape it. Suppression has never reversed the value of a new instrument. It has only delayed the industries capacity to adapt to it.

The practitioners who build the composition (AI for pattern, humans for the cost-bearing encounter, protocol design governing both) will shape what this field looks like in a decade. The rest will be defending a bottleneck that is no longer there.

This is a preview of an upcoming essay I am working on. To make sure you don’t miss it, subscribe below.