A systematic review of the most rigorous trials of EEG neurofeedback for internalizing disorders, a category that includes anxiety-related conditions, found only four reports that met its strict standard: randomized, double-blind, and sham-controlled. Together they enrolled 152 people. Three showed small to moderate effects favoring real feedback over sham, but every 95% confidence interval was wide enough to include no effect. A fourth had not posted results.
That is the state of the strongest evidence on whether neurofeedback works for the specific reason its advocates give. Anxiety affects an estimated 301 million people worldwide, and people are searching for options. Popularity, commercial availability, and a plausible mechanism do not establish clinical effectiveness.
What Neurofeedback Is and How It Works
A sensor records brain activity, software scores it against a target, and the person receives a sound or visual that moves when the signal shifts in the desired direction. Repeated sessions are meant to teach self-regulation, in the way a mirror helps someone adjust posture.
Neurofeedback is a form of biofeedback. Other forms train different signals. Heart rate variability (HRV) biofeedback, for example, trains the rhythm of the heartbeat, usually with paced breathing, and has a separate evidence base.
| Approach | Signal measured | Intended mechanism | Typical application | Evidence limitation |
|---|---|---|---|---|
| EEG neurofeedback (clinic) | Scalp brain waves, multiple electrodes | Learned control of brain rhythms | Anxiety, ADHD, PTSD, depression | Few sham-controlled trials |
| Consumer EEG headband | Few dry electrodes on forehead and ears | Feedback during meditation | Stress, relaxation | No sham controls in reviewed trials |
| fMRI neurofeedback | Brain blood flow in the scanner | Learned control of specific regions | Research, PTSD | Costly, scanner-based, mixed results |
| HRV biofeedback | Heart rhythm, breathing | Autonomic regulation | Stress, anxiety | Many studies lack strong controls |
Protocols differ in target signals, training environments, and intended outcomes, so products sold as neurofeedback are not interchangeable. Proposed mechanisms include learning, attention training, and regulation of arousal. Plausible mechanisms are not the same as mechanisms shown directly in clinical trials, and several reviews below found that participants often did not demonstrably change the targeted brain signal.
Claim One: “Neurofeedback Retrains the Brain, and That Reduces Anxiety”
The test. If the brain training itself drives improvement, real feedback should beat sham feedback, where the signal is random or from another person. Participants and raters must not know which they received.
The evidence. Under these conditions, evidence for specificity is thin. The four-trial, 152-person review reached an uncertain conclusion, and the author’s later doctoral work called the evidence for specificity tenuous at best. In ADHD, a field with more sham-controlled trials, several reported comparable improvement in real and sham groups.
Verdict: unproven. Improvement after neurofeedback is documented in some studies. Whether it comes from the targeted brain change is not established.
Claim Two: “It Works for Anxiety”
The test. Look at each condition separately, since anxiety disorders, stress, performance anxiety, and trauma-related disorders are different problems.
| Condition | Study design | Intervention | Comparator | Sample size | Symptom outcome | Follow-up | Risk of bias |
|---|---|---|---|---|---|---|---|
| Internalizing disorders (depression, PTSD, OCD, anxiety-related) | Systematic review of double-blind sham RCTs | EEG neurofeedback | Sham feedback | 4 reports, 152 participants | Small to moderate point estimates, wide intervals spanning no effect | Varied | Small samples |
| PTSD in adults | Meta-analysis of RCTs, 2025 | EEG neurofeedback | Passive controls such as waitlist | 5 trials | Moderate to large effect | Pre to post | Very low to low confidence |
| PTSD in adults | Same meta-analysis | fMRI neurofeedback | Sham | 2 trials | No improvement | Pre to post | Very low to low confidence |
| Stress and distress, consumer headbands | Meta-analysis, 2025 | Neurofeedback during mindfulness | Mostly mindfulness apps | 16 RCTs, 763 participants | g of negative 0.29, shrinking to negative 0.16 after strict correction | Up to 8 weeks | No sham controls |
| Stress and anxiety, HRV biofeedback | Meta-analysis, 2017 | HRV biofeedback | Control conditions | 24 studies, 484 participants | Between-group g of 0.83 | Varied | Authors call for better controls |
The same field produces opposite signals depending on the control. EEG neurofeedback looks strong against waitlist controls, which cannot account for expectation, attention, or therapist contact. The two sham-controlled fMRI neurofeedback trials showed no improvement. A separate 2024 meta-analysis of PTSD neurofeedback judged the quality of evidence moderate to high, which shows that reviewers disagree about how much weight the studies deserve.
Diagnosis matters here. A formal anxiety disorder, ordinary work stress, performance nerves, and trauma symptoms are different targets, and a study in one does not carry over to another. Wellness marketing often blurs them, using results from stressed students or healthy volunteers to imply benefit for people with clinical disorders.
Verdict: mixed and condition-specific. The evidence for generalized anxiety specifically is sparse, and nothing here supports a blanket statement that neurofeedback treats anxiety disorders.
Claim Three: “Consumer Brain Headbands Work for Stress”
The test. A 2025 meta-analysis in the Journal of Medical Internet Research pooled 16 randomized trials and 5 within-participant studies of consumer neurofeedback devices used during mindfulness.
The evidence. Samples were small, typically 30 to 50. Most trials compared the device with a mindfulness app, not a sham. Psychological distress fell modestly relative to controls (Hedges g of negative 0.29, or negative 0.16 under strict correction for selective reporting). There was no evidence of better mindfulness, cognition, or physiological outcomes, and participants did not show clear modulation of the brain target. No study used sham neurofeedback, and adverse events were assessed in only 2 of 21 studies. No trial ran longer than 8 weeks.
The authors considered a placebo-like “neurosuggestion” effect plausible: belief that a brain device is helping can itself reduce distress. That does not make the effect worthless, but it separates symptom relief from proof that the brain was retrained.
Expectation deserves a plain explanation. A session combines a device, a ritual, regular attention, and often a supportive practitioner. Each can lower distress without any change in brain activity. Placebo effects are real effects on how people feel, and some researchers consider them a legitimate part of care. They still matter when judging claims, because a person paying for brain retraining deserves to know what the data can and cannot show.
Verdict: small effect, unclear reason. Devices cost roughly $100 to $200 according to the authors, far less than therapy courses. Low cost matters, and so does the absence of sham-controlled evidence.
Claim Four: “It Is Better Than, or Equal to, Standard Treatment”
The test. A fair comparison needs head-to-head trials in the same condition with the same outcome.
The evidence. No such trial for neurofeedback was identified in this review. By contrast, an 8-week randomized trial of 276 adults with anxiety disorders found mindfulness-based stress reduction (MBSR) noninferior to the antidepressant escitalopram, with mean Clinical Global Impression of Severity scores of 3.09 in both groups and benefits persisting at 24 weeks. Established psychotherapies, notably cognitive behavioral therapy, have a deep evidence base that this guide did not re-review.
| Intervention | Evidence base | Treatment format | Limitations | Clinical considerations |
|---|---|---|---|---|
| Cognitive behavioral therapy | Extensive randomized evidence, recognized first-line | Weekly sessions, in person or digital | Access, skills practice required | Established first choice for many |
| Medication, such as escitalopram | Extensive evidence | Daily dosing | Side effects, adherence | Do not stop without clinician advice |
| MBSR | Noninferior to escitalopram in 276 adults | 8-week group course | Time commitment | Alternative where available |
| HRV biofeedback | Meta-analysis, large effect, weak controls | Guided breathing practice | Study quality | Low-risk adjunct candidate |
| EEG neurofeedback | Mixed, few sham-controlled trials | Many sessions with a provider | Cost, uncertain specificity | Not a replacement for established care |
Verdict: not shown. Equivalence should not be implied without direct comparative evidence.
Claim Five: “It Is Risk-Free”
The test. Safety needs active monitoring.
The evidence. Adverse events are often not tracked. In the consumer-device meta-analysis, only 2 of 21 studies monitored for them, and neither reported any. Absence of reports is not proof of safety. Because side effects are rarely tracked, how often fatigue, frustration, or other discomfort occurs is unknown. The bigger risk is opportunity cost: delay in starting an effective treatment while sessions continue, and sunk-cost reluctance to stop when progress is unclear.
Verdict: low direct risk, real indirect risk. Prescribed medication should not be stopped because of neurofeedback.
What a Session Involves
Assessment usually comes first, including history, symptom scales, and sometimes a baseline recording. Sensors are placed on the scalp. The person watches a screen or listens to sounds that respond to their brain signal, often while relaxed. Sessions are reviewed and protocols adjusted.
Session count, duration, and cost vary by provider and protocol, and no standard dose is established for anxiety. Qualifications also vary. Some providers are licensed clinical psychologists or physicians, while others hold certificates in neurofeedback only.
Costs and Limits
Total cost depends on session fees, number of sessions, and any home equipment. Insurance coverage is inconsistent and should be checked in advance. Commercial protocols are rarely comparable, so published results for one system do not transfer to another.
How to Evaluate a Provider
| Question | Evidence to request | Why it matters |
|---|---|---|
| What are your credentials? | License, training, supervision | Diagnosis and safety need clinical training |
| Which protocol, and for what condition? | Protocol name and studies for that condition | Evidence is protocol- and condition-specific |
| How do you measure progress? | Symptom scales at baseline and intervals | Prevents open-ended treatment |
| What happens if I do not improve? | Stopping rule and referral plan | Avoids sunk-cost drift |
| What are the total costs? | Written fee schedule, session estimate | Costs add up |
| What are the risks and alternatives? | Discussion of CBT, medication, MBSR | Informed choice |
A responsible provider welcomes questions, defines success in advance, and coordinates with the patient’s prescriber or therapist. Be cautious about promises of cure, guaranteed results, or advice to stop medication.
The Future of Neurofeedback Research
The field needs larger sham-controlled trials, objective outcome measures, evidence that participants actually learn to change the target signal, and longer follow-up. Direct comparisons with established treatments are needed. Personalized protocols and better devices may help, and they must be tested the same way. Emerging research should be distinguished from accepted clinical practice.
Conclusion: Judge Each Protocol and Condition Separately
Neurofeedback is a group of interventions whose effectiveness must be assessed by protocol and condition. The current evidence shows possible benefit in some studies, uncertainty about why, and a thin base of sham-controlled trials for anxiety. Stronger comparative studies and longer follow-up are needed.
Treatment decisions should reflect reliable evidence, individual needs, and the availability of established care. People who want to try neurofeedback can do so as an addition to, not a substitute for, evidence-based treatment.
Medical Disclaimer: This article is educational and does not replace diagnosis or treatment by a qualified mental health professional. People experiencing significant anxiety should seek professional assessment, and prescribed treatment should not be stopped without clinical advice. Anyone in crisis or thinking about self-harm should contact local emergency services or a crisis line immediately.
FAQ
Q: What is neurofeedback therapy?
A: It is a form of biofeedback in which a person sees or hears a real-time signal of their brain activity, commonly EEG, and practices changing it.
Q: Does neurofeedback work for anxiety?
A: Evidence is mixed. Sham-controlled trials are few and small, and benefits may reflect nonspecific factors such as expectation.
Q: Is EEG neurofeedback different from biofeedback?
A: EEG neurofeedback is one kind of biofeedback. HRV biofeedback trains heart rhythm and has a separate evidence base.
Q: How many neurofeedback sessions may be involved?
A: It varies by provider and protocol, and no standard number is established for anxiety.
Q: Is neurofeedback better than cognitive behavioral therapy?
A: No direct evidence shows that. CBT has a much larger evidence base for anxiety.
Q: Are there risks associated with neurofeedback?
A: Direct risks appear low, though adverse events are rarely monitored. The main risk is delaying effective treatment.
Q: How can patients evaluate a neurofeedback provider?
A: Check credentials, protocol evidence for the specific condition, outcome measures, costs, and a plan if there is no improvement.
Q: Can neurofeedback replace conventional mental health treatment?
A: No. It should not replace established treatment or prescribed medication without clinical advice.