AI News Page

FIELD GUIDE · THE 3 A.M. THERAPIST

SHOULD AI BE YOUR THERAPIST?

It's cheap, awake at 3 a.m., and never looks shocked. Real perks. Also exactly why people confuse always-available with actual care.

UPDATED 2026-09-22

TL;DR

THE STORY SO FAR

01

PATIENTS ARE ALREADY BRINGING CHATBOTS INTO THERAPY

APA surveyed 1,200+ licensed U.S. psychologists in 2026: 77% had patients discussing AI use, and over a third said patients were using AI as an extra mental-health provider. Mainstream enough to change how therapy works.

02

APA DOES NOT RECOMMEND GENERAL-PURPOSE CHATBOTS AS A REPLACEMENT

APA's health advisory says general-purpose chatbots shouldn't replace qualified providers — while allowing that narrower supportive uses can help sometimes. The serious position is backup, not replacement.

03

CLINICIANS REPORT BOTH HELP AND DEPENDENCY

Same 2026 survey: some patients did well, but 36% of psychologists noticed chatbot dependency and 15% saw distorted thinking or delusions tied to chatbot use. Not the same for everyone.

04

HEAVY OR ATTACHMENT-PRONE USE CAN CORRELATE WITH WORSE OUTCOMES

An OpenAI–MIT study found results depended on how people used it: long daily sessions and stronger attachment were linked to worse psychosocial outcomes. The pattern of use matters as much as the use itself.

KEEP READING

MORE QUESTIONS

SHOW YOUR WORK

SOURCES

The demand is real. The evidence for replacing clinicians is not.

  1. PATIENTS ARE ALREADY BRINGING CHATBOTS INTO THERAPYPatients are bringing AI to therapyAmerican Psychological Association
  2. APA DOES NOT RECOMMEND GENERAL-PURPOSE CHATBOTS AS A REPLACEMENTHealth advisory: Use of generative AI chatbots and wellness applications for mental healthAmerican Psychological Association
  3. SIMULATED EMPATHY IS NOT CLINICAL CAREDiscussing AI use in therapyAmerican Psychological Association
  4. HEAVY OR ATTACHMENT-PRONE USE CAN CORRELATE WITH WORSE OUTCOMESEarly methods for studying affective use and emotional well-being on ChatGPTOpenAI and MIT Media Lab