top of page
Search

It Wasn't the Advice That Kept Them Coming Back: What 1,187 People Taught Me About Empathy and Showing Up

Nov 14
6 min read

The gym membership is still in your email. The app you downloaded is gone from your phone. The journal has three days filled in, then a long stretch of blank pages.

If that sounds familiar, I want to say this gently: drifting away from a helpful habit does not make you lazy or unmotivated. Sometimes the thing that could help is also the thing that feels hardest to return to.

A 2026 trial of 1,187 adults asked a narrow question about a self-guided online CBT program: could automated feedback help people keep showing up? Its most striking finding was not that the program’s scores improved more. They did not. It was that an empathic response was linked to people staying with the exercises, while practical advice was not significantly linked to staying.

That difference feels worth sitting with.

What participants actually did

The participants were adults aged 20 to 60 in Japan, with mostly mild to moderate depressive symptoms. Their average age was 43.5, most were employed, and slightly more than half were men. Only about one in ten was already using mental health services. The program was delivered in Japanese, and participants completed six weekly sessions.

Each session included a short video about CBT ideas such as problem solving and behavioral activation, followed by an exercise. Participants wrote about a situation, an automatic thought, and a feeling.

One group used the self-guided program with automated feedback. Another used the same program without it, and a third was assigned to a waitlist. The feedback came through text and voice from an animated character. One function responded empathically to what someone had written. Another offered practical suggestions, such as pointing out when a feeling had been entered in the space for a thought.

This was a self-guided program with no therapist involved. The animated character was not a therapist or counselor. The study tested whether adding feedback to a program affected engagement and outcomes.

The honest result: no added improvement in depression scores

Here is the part I do not want to overstate: the AI-augmented program did not lead to significantly different average depression scores compared with the other groups at week seven or at three months. Both active program groups’ scores decreased over time, but the study did not find a significant between-group difference in its primary outcome.

So, if the question is whether this automated feedback made depression scores better, the honest answer is that this trial did not show that.

There was an exploratory finding among a more restricted group of participants who completed at least three of the six sessions. At the three-month follow-up, only the AI group had a significantly lower proportion of participants scoring 10 or above on the PHQ-9 compared with the waitlist group. The authors describe this as exploratory and call for more research. It is not proof of a lasting benefit.

The clearer result was about participation. People in the AI group had higher odds of completing at least one exercise each week than people in the program without AI feedback. Participation began at about half of the participants in week one and fell over time. By week six, it was around 30 percent in the no-AI group and in the low 40s in the AI group.

That is not the same as feeling better. But it asks an important question: what helps someone come back for another session?

Empathy, not instructions

In an exploratory analysis, participants whose empathic feedback function was activated during the first week were much more likely to keep doing exercises in the following weeks. The reported odds ratio was 9.99, with a 95 percent confidence interval of 5.80 to 17.21. Activation of the advisory function was not significantly linked to staying.

I want to hold that carefully. This was an exploratory finding from one trial. It does not prove that empathy caused people to continue, or that advice is unhelpful. It suggests that being responded to with warmth may matter in a way that a correction or instruction alone does not.

Two adults in a warm conversation, one leaning in and listening attentively

Advice can be useful. A clear next step can make a task feel less overwhelming. But advice often lands best after someone feels heard, respected, and met where they are. Without that sense of connection, even good instructions can feel like one more thing to get right.

If you have downloaded an app, started a workbook, or begun journaling and then drifted away, that drift is information, not a character flaw. Maybe the task did not fit your life. Maybe you needed more support than a program could offer. Maybe the first step felt too big, or the response you got felt too impersonal. We can be curious about that without shaming ourselves.

What I want a therapy room to offer

As a Talk to Heal clinician, I want our work to begin with listening, not a ready-made to-do list. I want to understand what this season of life is asking of you, what has helped you get through it so far, and what kind of support feels respectful and manageable.

Therapy can include practical tools and small steps. But we build them together. I can help you navigate what feels stuck while making room for your strengths, your pace, and your experience. A safe, supportive, and respectful relationship is not an extra feature around the work. It is part of how we do the work.

The study’s automated feedback function is not a substitute for that relationship, and the findings do not tell us what happens in therapy with a real clinician. They do remind me that a person may need more than a correct instruction. They may need a place where they can say, “I stopped,” and be met with curiosity rather than judgment.

An adult in their thirties sitting in soft sunlight with a phone face down nearby, looking thoughtful and settled

Hold this study loosely

These participants were mostly working adults in their forties in Japan, recruited through a market-research panel with a strong affinity for digital tools. They had mostly mild symptoms, so the results may not generalize to people with different experiences or greater levels of distress. The program was entirely in Japanese, and this was one trial in one country.

Dropout was high. Only 37.6 percent of the AI group and 33.8 percent of the other program group reached the three-session threshold used for the exploratory analysis. About half of each group had already stopped by the very first session. Lower adherence was associated with being male, older, and employed. The authors describe the study as exploratory and say it needs replication. Most importantly for this article’s focus, the program was self-guided and had no therapist involved, so it says nothing about therapy with a real clinician.

I invite you to take what fits and leave what does not. One trial cannot explain why any one person stops or continues.

A few gentle ways to begin again

Notice what you actually return to. Not what you think you should want. Which kind of support feels easier to come back to: a brief conversation, a few lines in a notebook, or one small action alongside someone you trust?

Tell someone what you need. You might try, “I’m not looking for advice yet. Could you listen for a few minutes?” Naming the kind of support you want can make it easier for another person to meet you.

Let one person know you stopped. You do not need to catch up or explain it perfectly. You can simply say, “I started and then fell away. I’d like help figuring out what would feel manageable now.”

A close view of a notebook and tea on a table, with another person’s hand nearby in a gentle gesture of support

Support in Georgia

If you would like to explore therapy, I welcome you to start with a free 15-minute consultation or book an appointment online. Talk to Heal Counseling Center offers therapy in many different languages and accepts insurance including UHC, Aetna, Cigna, and many others.

Care is provided only in the State of Georgia. For online sessions, you must be physically located in Georgia at the time of your session. To ask a question or get in touch, call 404-369-3838.

This is not a crisis service. If you are in immediate danger, call or text 988, or call 911.

You do not have to prove your motivation before asking for support. You can begin from where you are, including from the blank pages or the unfinished program. I hope you find a kind of support that makes it feel possible to return.

Source: So M, Sekizawa Y, Hashimoto S, Kashimura M, Yamakage H, Watanabe N. “Effect of AI-Based Natural Language Feedback on Engagement and Clinical Outcomes in Fully Self-Guided Internet-Based Cognitive Behavioral Therapy for Depression: 3-Arm Randomized Controlled Trial.” Journal of Medical Internet Research, 2026 Jan 5;28:e76902. doi:10.2196/76902. PMCID PMC12817041. Read the open-access study.

 
 
 

Comments


bottom of page