Back to Bigger Picture

BIGGER PICTURE Better care

A therapy chatbot was tested for eight weeks. What changed?

Dartmouth tested whether a purpose-built, clinically supervised chatbot could improve measured mental-health symptoms. Participants received four weeks of prompted use, followed by four more weeks when they could continue using it voluntarily.

AI-generated illustration: A structured text conversation represents the chatbot studied in a controlled trial.
A structured text conversation represents the chatbot studied in a controlled trial.AI-generated illustration of the idea, not a photograph of the actual people, equipment or event.
Randomized groups
106 with access; 104 on a waitlist
Access period
Four weeks with prompts; four further weeks of optional use
Depression change in users
51% average reduction from baseline after eight weeks

How the idea works

  1. Describe the difficult moment

    Participants type concerns or respond to prompts.

  2. Generate focused dialogue

    A purpose-built model draws on therapeutic conversation patterns.

  3. Track symptoms under oversight

    Researchers compare measured symptom change with a waitlist.

A conversation outside an appointment

Participants could type into a smartphone app when they wanted support, as well as respond to prompts. Therabot used open-ended dialogue developed around therapeutic practices. This made it possible to return to a concern when it arose, rather than waiting for a scheduled study interaction.

Sources: [1]

An actual comparison, with a limited comparator

The randomized study assigned 106 participants to Therabot and 104 to a waitlist. Prompted use lasted four weeks; optional access continued for four further weeks. It included depression, anxiety and eating-disorder-risk groups. Researchers followed symptoms rather than treating satisfaction with the chatbot as sufficient evidence that it helped.

Sources: [1][2]

Read the improvement figure correctly

The Dartmouth report describes an average 51% reduction in depression symptoms among Therabot users after eight weeks. This is change from their own starting scores, not a 51-percentage-point advantage over a therapist. The trial found improvements relative to the waitlist; it did not compare Therabot directly with professional psychotherapy.

Sources: [1][2]

Clinical oversight belongs to this result

The study team stressed the continuing need for clinician oversight. Its result applies to this specifically developed system and monitored research setting. It offers a reason to investigate accessible, timely support further, not evidence that any general chatbot can safely deliver mental-health treatment on its own.

Sources: [1][2]

The original sources

  1. Dartmouth CollegeFirst Therapy Chatbot Trial Yields Mental Health Benefits
  2. NEJM AIRandomized Trial of a Generative AI Chatbot for Mental Health Treatment

What to explore next

A warning before blood pressure falls, with a plan for what to doThe kitchen bin can explain tomorrow’s breakfast
Explore all 99 applications