← All health signals
Wellness turn

Machine Therapy

The counsellor that never tires and never judges normalises talking things through with something that isn’t a person.

What is it?

A machine that never tires, never bills and never judges is absorbing the conversations people once saved for a therapist, or never had at all. Clinical trials show real gains. Whether that comfort substitutes for human contact, or rehearses it, stays open.

Why now?

Loneliness scales faster than the clinician supply meant to answer it — a widening gap between people who want to talk and the trained hours available to listen. AI collapses that shortage to nothing: infinite patience, zero wait. A structural gap becomes a software feature.

Who’s building it?

Consumer AI companies and wellness startups build the tools; clinicians and regulators are only now setting the guardrails.

Companion-app makersClinical trial teamsWellness platformsLicensing boardsInsurers · employers
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — AI wellness companies and the model providers beneath them.
Buyers — the isolated, the anxious, and anyone awake at 3am wanting someone to talk to.
Gatekeepers — clinical bodies, app-store platforms, and regulators weighing where oversight begins.
What’s at stake
One serious harm story could shut the door on a decade of goodwill — trust compounds slowly and breaks fast.
Data intimacy is the moat, and also the liability.
Clinician endorsement could unlock adoption at clinical scale.
What’s unsettled
Whether AI support substitutes for therapy, or simply extends its reach.
Who is liable when a chatbot’s advice causes real harm.
Whether clinical-grade results replicate at scale outside a controlled trial with careful screening.
02 · How the demand matureda progression, not a timeline
Access
People try it privately, for a need too small to book.
Habit
It becomes the default outlet — always open, never scheduling a return visit.
Clinical uptake
Trial evidence and clinician comfort turn a private habit into something worth recommending.
Category
Enough paying, returning users to justify real safety standards, real liability, real rivals.
03 · Where the market goesthe forks, not a prediction
→Converge or fragment — decided by whether insurers and clinical bodies set one accredited standard, or a fragmented market of unregulated apps persists.
→Clinicians slowly shift from gatekeepers to supervisors, curating which tools are safe enough to recommend.
→Employers and insurers quietly fold AI support into benefits, one liability case away from a reckoning.
→Or the profession absorbs it fully, and machine therapy becomes a covered first line.
The open question is whether relief here transfers into real-world contact, or the conversation with the machine quietly becomes the whole relationship.
04 · The Synthetic Companionship deskadjacent markets, moving now
The Loneliness Market
Connection sold like any other good.
27% of US adult internet users have significant emotional or social exchanges with AI.
Elon University
The Digital Afterlife
Grief turned into a permanent, paying subscription service.
6+ griefbot platforms now serve millions of users (2026).
The Week
Emotional Outsourcing
Even apologies now arrive pre-written by the machine.
AI-written apologies rated insincere, lazy vs human ones (2026).
The Conversation
05 · The Wire
01
51% drop in depression, 31% in anxiety over 8 weeks.
Dartmouth ↗
02
35% of psychologists see patients using AI help.
MedicalXpress ↗
03
27% of US adults use chatbots for personal or emotional queries
The Washington Post ↗
The panel on this signal
They agreed on
–The comfort and the risk are the same feature — always-on availability that a busy clinician cannot match.
–The clinical evidence for gains is genuinely strong.
–Oversight will land above the builders, set by regulators and clinical bodies.
The clinical specialist disagreed on
–External validity is the concern: a supervised, time-limited study with excluded high-risk participants cannot forecast outcomes once the same tool reaches an unscreened general population.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.