Using AI for therapy? Don't - it's bad for your mental health, APA warns

Yaroslav Kushta / Moment / Getty Images

Follow ZDNET: Add america arsenic a preferred source connected Google.

ZDNET's cardinal takeaways

• Consumer AI chatbots cannot regenerate intelligence wellness professionals.
• Despite this, radical progressively usage it for intelligence wellness support.
• The APA outlines AI's dangers and recommendations to code it.

Therapy mightiness beryllium costly and inaccessible, portion galore AI chatbots are escaped and readily available. But that doesn't mean the caller exertion tin oregon should regenerate intelligence wellness professionals -- oregon afloat code the intelligence wellness crisis, according to a caller advisory published Thursday by the American Psychological Association.

Also: Is ChatGPT Plus inactive worthy $20? How it compares to the Free and Pro plans

The advisory outlines recommendations for the public's usage and over-reliance connected consumer-facing chatbots. It underscores the wide nationalist and susceptible populations' increasing usage of uncertified, consumer-facing AI chatbots and however they're poorly designed to code users' intelligence wellness needs. 

Largest providers of intelligence health

Recent surveys show that 1 of the largest providers of intelligence wellness enactment successful the state close present is AI chatbots similar ChatGPT, Claude, and Copilot. It besides follows respective high-profile incidents involving chatbots' mishandling of radical experiencing intelligence wellness episodes. 

In April, a teenage lad died by termination after talking with ChatGPT astir his feelings and ideations. His household is suing OpenAI. Several akin lawsuits with different AI companies are ongoing.

Also: ChatGPT lets parents restrict contented and features for teens present - here's how

(Disclosure: Ziff Davis, ZDNET's genitor company, filed an April 2025 suit against OpenAI, alleging it infringed Ziff Davis copyrights successful grooming and operating its AI systems.)

Through validation and amplification of unhealthy ideas oregon behaviors, immoderate of an AI chatbot's tendencies tin really aggravate a person's intelligence illness, the APA says successful the advisory.

Not reliable attraction resources

The APA outlines respective recommendations for interacting with consumer-facing AI chatbots. The chatbots are not reliable psychotherapy oregon intelligence attraction resources, the APA says. OpenAI CEO Sam Altman has said the same. 

In an interrogation with podcaster Theo Von, Altman advised against sharing delicate idiosyncratic accusation with chatbots similar OpenAI's ain ChatGPT. He besides advocated for chatbot conversations to beryllium protected by akin protocols that doctors and therapists ahere to, though Altman mightiness beryllium much motivated by legally protecting his company.

The advisory outlined recommendations for preventing dependencies with chatbots whose extremity is to support "maximum engagement" with a user, the APA says, alternatively of achieving a steadfast outcome. 

"These characteristics tin make a unsafe feedback loop. GenAIs typically trust connected LLMs trained to beryllium agreeable and validate idiosyncratic input (i.e., sycophancy bias) which, portion pleasant, tin beryllium therapeutically harmful, reinforcing confirmation bias, cognitive distortions, oregon avoiding indispensable challenges," constitute the authors of the advisory.

Also: ChatGPT volition verify your property soon, successful effort to support teen users

By creating a mendacious consciousness of therapeutic alliance, being trained connected clinically unvalidated accusation crossed the internet, incompletely assessing intelligence health, and poorly handling a idiosyncratic successful crisis, the APA says these consumer-facing chatbots airs a information to those experiencing a intelligence wellness episode.

"Many GenAI chatbots are designed to validate and hold with users' expressed views (i.e., beryllium sycophantic), whereas qualified intelligence wellness providers are trained to modulate their interactions -- supporting and challenging -- successful work of a patient's champion interest," the authors write.

The onus is connected AI companies

The APA puts the onus connected companies processing these bots to forestall unhealthy relationships with users, support their data, prioritize privacy, forestall misrepresentation and misinformation, and make safeguards for susceptible populations. 

Policy makers and stakeholders should besides promote AI and integer literacy education, and prioritize backing for technological probe of generative AI chatbots and wellness apps, the APA says. 

Also: If your kid uses ChatGPT successful distress, OpenAI volition notify you now

Ultimately, the APA urges the deprioritization of AI to code systemic issues of the intelligence wellness crisis.

"While AI presents immense imaginable to assistance code these issues," the APA authors write, "for instance, by enhancing diagnostic precision, expanding entree to care, and alleviating administrative tasks, this committedness indispensable not distract from the urgent request to hole our foundational systems of care."