By Stuart Kerr, Technology Correspondent, LiveAIWire

Can AI therapies heal trauma, or do they just make it easier for patients to approach it in the first place? The strongest evidence so far comes from BraveMind, a virtual reality exposure system built at the University of Southern California’s Institute for Creative Technologies. In the largest randomized clinical trial conducted to date, BraveMind was found to be equivalent to Prolonged Exposure, the gold-standard evidence-based treatment for PTSD, and produced better outcomes in patients with co-occurring major depression. Seventy-six percent of incoming patients in that trial said they preferred VR therapy over the traditional talk-based version.

BraveMind does not use AI to write or reprocess a patient’s story. It uses simulation technology to recreate the sensory conditions of a traumatic memory, visual scenes, directional 3D audio, vibration, and smell, while a trained clinician guides the exposure session in real time through a separate control interface. According to the VA’s own innovation tracker, the system has been deployed across 28 VA medical facilities and more than 170 clinical sites overall, and has been adapted for survivors of military sexual trauma and, most recently, Ukrainians displaced by war.

Can AI Therapies Heal Trauma? What BraveMind’s Trial Actually Found

Exposure therapy has been an evidence-based PTSD treatment for decades, built on the principle that a patient who safely revisits a traumatic memory, guided by a therapist, gradually loses the automatic fear response attached to it. The limitation has always been that imagining a scene vividly enough to trigger genuine therapeutic exposure is difficult for many patients, particularly those with comorbid depression, who often struggle to generate detailed mental imagery on demand. BraveMind’s virtual environment solves that specific problem by supplying the sensory detail a patient’s own imagination might not reliably produce.

That is the mechanism behind the finding that BraveMind produced better outcomes than standard Prolonged Exposure specifically in patients with co-occurring depression. It is not that virtual reality is inherently more powerful than imagination. It is that some patients cannot reliably access the imaginative detail exposure therapy depends on, and a simulated environment removes that barrier. The distinction matters because it defines who benefits most: patients who struggle with vivid recall, not necessarily every PTSD patient equally, and the treatment remains clinician-led throughout rather than automated.

The Chatbot Side: What Therabot Showed, and What It Didn’t

A separate and much-discussed strand of AI mental health research involves conversational chatbots rather than VR simulation. Therabot, a generative AI chatbot developed at Dartmouth College, was tested in a randomized controlled trial published in the New England Journal of Medicine AI in 2025. Over four weeks, 106 participants using Therabot showed average symptom reductions of 51 percent for major depressive disorder and 31 percent for generalized anxiety disorder, compared with a waitlist control group of 104 participants.

What the Therabot trial did not test is PTSD or trauma specifically. Its three studied conditions were major depressive disorder, generalized anxiety disorder, and clinical high risk for feeding and eating disorders. Senior researcher Nicholas Jacobson said the team had not expected participants to bond with the software the way they did, noting some initiated conversations in the middle of the night. First author Michael Heinz was explicit about the limits: no generative AI agent is ready to operate fully autonomously in mental health where there is a very wide range of high-risk scenarios it might encounter. Applying Therabot’s results to trauma treatment specifically would overstate what the trial actually measured.

What This Means If You or Someone You Love Has PTSD

The practical distinction for anyone evaluating AI-assisted trauma treatment is between two genuinely different technologies wearing a similar label. VR exposure systems like BraveMind have trial evidence specific to PTSD, delivered under the supervision of a trained clinician, and are already integrated into VA and military treatment programs rather than offered as a standalone consumer product. General-purpose AI chatbots, even therapeutically fine-tuned ones like Therabot, have real evidence for depression and anxiety but have not been validated for trauma treatment and are not designed to replace a clinician managing a PTSD case.

LiveAIWire’s broader look at whether AI therapy works found that chatbot tools are most effective for people with moderate depression or anxiety symptoms and no acute crisis indicators, and found serious inconsistencies in how reliably popular chatbots recognize suicide risk during testing. That caution applies with even more force to trauma treatment, where a poorly calibrated tool could retraumatize rather than help a patient working through a specific memory, which is precisely why BraveMind keeps a trained clinician directly in the loop for every session.

Where the Limits Still Are

Neither BraveMind nor Therabot operates without a human clinician involved, and that is not an incidental detail. BraveMind requires a specially trained therapist controlling the exposure scenario and in constant audio contact with the patient throughout the session. LiveAIWire’s analysis of AI positioned as a therapist found that the strongest evidence for AI mental health tools consistently involves augmenting a clinician’s work rather than replacing it, and that the risk of treating AI as a full substitute for human judgment grows as the clinical stakes rise. Trauma treatment sits near the top of that risk scale.

Detecting who might benefit from these tools before symptoms become severe is itself an area of active research. LiveAIWire’s coverage of AI depression detection from speech patterns found that AI screening tools can identify warning signs earlier than standard clinical questionnaires in some studies, though performance drops outside the relatively narrow populations most models have been trained and tested on. The same caution applies to trauma screening: promising research results are not the same as a validated, equitable clinical tool ready for widespread deployment.

So, Can AI Therapies Heal Trauma?

The honest answer to whether AI therapies heal trauma is that AI-adjacent tools can meaningfully support treatment, but only the specific, clinically supervised versions with trial evidence behind them, and only as an enhancement to a therapist’s work rather than a replacement for it. BraveMind’s trial results are genuinely strong and specific to PTSD, gathered over years of trials across VA, DoD, and university clinical sites. Therabot’s results are genuinely strong and specific to depression, anxiety, and eating disorder risk, not trauma, and its own researchers have said as much.

Conflating the two, or assuming any AI-branded mental health app carries the same evidence behind it, is the mistake that matters most for anyone deciding whether to trust one of these tools with their own recovery. Whether AI therapies heal trauma turns out to be the wrong framing entirely; the more useful question is which specific, evidence-tested tool fits which specific condition. The technology that can plausibly claim to heal trauma right now is the one built specifically for it, tested specifically on it, and delivered by a human clinician who never leaves the room.

About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.