Skip to content
MyFreud

By , Editorial Team Published · Updated

Online CBT helped a third of people, and that is the number

A 329-person trial of internet CBT for social anxiety in Hong Kong found reliable improvement in 34 to 40 percent, against 11 percent on the waitlist.

3 min read

Pop-art illustration of a woman working at a laptop beside a window.

Key takeaways

  • Reliable improvement reached 34 percent on web and 40 percent on app, against 11 percent on the waitlist. Most participants did not reliably improve.
  • Reporting the proportion who got better, rather than only an average change, is more informative and less common than it should be.
  • Web and app formats performed comparably, so the delivery channel mattered less than getting the treatment at all.
  • Gains held at three and six months, which is longer follow-up than most digital trials manage.
  • The comparison was a waitlist, so this measures internet CBT against nothing rather than against a therapist.

Most trials report how much scores moved on average. This one reports how many people actually got better, which is a different and more useful number, and it is lower than the average would suggest. [pubmed-social-anxiety-jul13-2026-source]

What the researchers did

Three hundred and twenty-nine Hong Kong Chinese adults with social anxiety disorder, average age 30, were randomised into three groups:

  • Web-based internet CBT with virtual reality exposure (117)
  • App-based internet CBT with virtual reality exposure (111)
  • Waitlist control (101)

The programme, “Ease Online”, ran 14 weeks. Assessments were taken before, after, and at three and six months.

What they found

Both treatment groups beat the waitlist on the primary and secondary measures, and the two formats performed comparably. Improvements held at three and six months.

Then the more informative statistics:

Reliable improvement: 34% on web, 40% on app, against 11% on the waitlist.

Remission: 15% and 23%, against a lower waitlist figure.

Why those numbers are the useful ones

A statistically significant average improvement can be produced by a large number of people improving slightly, or a few improving greatly, and it does not tell an individual what their chances are.

Reliable improvement counts people. It asks how many changed by more than measurement error can explain. Here, roughly a third to two fifths did, against about one in nine of those given nothing.

Set the two side by side and the honest summary is that internet CBT produced reliable improvement in around a quarter more people than would have improved anyway. That is a genuine effect for a programme needing no therapist time, and it means most participants did not reliably improve.

Both halves of that sentence belong in any accurate description, and the second half is the one that usually goes missing.

What the comparison does and does not cover

The comparison was a waitlist, so this is internet CBT against nothing.

Face-to-face cognitive therapy for social anxiety, particularly the model developed by one of this paper’s own authors, generally reports considerably higher recovery rates. Nothing here suggests the online version matches it.

What it offers instead is availability. Social anxiety is common, under-treated, and its central symptom makes people less likely to seek face-to-face help. A programme with no waiting list and no room to walk into reaches people who would otherwise receive nothing, and a third of them improving is a real gain over zero.

Would a self-directed programme suit you?

Internet CBT works better for some people than others. Tick what is true.

0 of 6 ticked

The follow-up length is a quiet strength

Most digital mental health trials stop measuring when the programme stops. This one followed people to six months and the gains held.

That matters because the usual pattern in this field is a strong post-treatment result that quietly fades, and because most published trials would not have detected the fade if it had happened. Reporting a maintained effect at six months is a stronger claim than a large one at week 14.

If social anxiety is affecting you

Our social anxiety guide covers the condition and the established treatments, and Gen Z and social anxiety covers why rates appear to be rising in younger adults.

The source

These findings are drawn from “Effectiveness of internet-based cognitive behavioral therapy with virtual reality exposure therapy for social anxiety disorder: A randomized controlled trial in Hong Kong” (Pan JY, Thew GR, Clark DM, 2025), published in the Journal of affective disorders. Read the full study on PubMed.

Frequently asked questions

What does reliable improvement mean?

A change large enough to be confident it is not just measurement noise, calculated from the known reliability of the questionnaire. It is a stricter and more meaningful standard than a statistically significant group difference, because it counts individuals who genuinely got better rather than averaging everyone together.

Is a third of people improving a good result?

It is a real one, and it is worth reading alongside the 11 percent who improved on the waitlist. Roughly a quarter more people improved than would have anyway. That is meaningful for a self-directed programme requiring no therapist time, and it also means most participants did not reliably improve, which is the part usually left out.

Did the app beat the website?

Not meaningfully. The two formats performed comparably, and the numerical gap between 34 and 40 percent is well within what a trial this size could produce by chance. The useful conclusion is that the delivery channel is not the important variable, which is worth knowing before anyone invests heavily in one over the other.

How does this compare with seeing a therapist?

It does not, because that comparison was not made. Both arms were compared against a waitlist. Face-to-face CBT for social anxiety, particularly the cognitive therapy model, generally reports higher recovery rates than these. What internet CBT offers is availability, not equivalence.

Why does it matter that the trial was in Hong Kong?

Because most CBT evidence comes from Western populations, and social anxiety involves norms about self-presentation, deference and face that vary by culture. A programme validated in a Chinese community adds evidence where there was little, which is the stated reason the researchers ran it there.

References

  1. 1.Pan JY, Thew GR, Clark DM ( 2025). Effectiveness of internet-based cognitive behavioral therapy with virtual reality exposure therapy for social anxiety disorder: A randomized controlled trial in Hong Kong. Journal of affective disorders. pubmed.ncbi.nlm.nih.gov . doi:10.1016/j.jad.2025.04.100

Related articles

Social Anxiety Disorder: Signs, Causes, Treatment

A 2026 guide to social anxiety disorder: how it differs from shyness, why it develops, how it is diagnosed, and which therapies and medications help most.

Social Anxiety vs Introversion: How to Tell Them Apart

One is a preference about how you spend energy and one is a fear of being judged. Here is the test that separates them, and why the mix-up delays treatment.

CBT for Social Anxiety: What the Treatment Actually Involves

CBT for social anxiety is not about thinking positively. It works by dropping the safety behaviours that feel like they are helping, and testing what happens.

Safety Behaviours: Why Coping Keeps Social Anxiety Going

The things you do to get through a social situation are the reason it never gets easier. An experiment dropping them beat exposure on its own, in one session.