Neurofeedback's evidence for tinnitus is small and mixed: an uncontrolled study reported modestly lower distress for six months, a 94-patient randomised trial found no distress advantage over a minimal intervention, and a scanner-based trial's win over CBT compared unequal treatments.
It is a different approach from TMS, which stimulates the brain from outside. Neurofeedback instead trains you to shift your own brainwave patterns, using real-time EEG feedback as a guide.
How it works
Neurofeedback measures your brain's electrical activity (EEG) in real time and displays it back to you, typically as a game or visual cue that responds to specific brainwave patterns. For tinnitus, the target is usually the alpha/delta ratio — increasing alpha waves (associated with relaxed, calm attention) relative to delta waves, based on research linking altered activity in these bands to tinnitus perception and distress. Over repeated sessions, patients learn to shift this ratio somewhat voluntarily, similar in principle to biofeedback but targeting brain activity specifically rather than muscle tension or skin conductance.
The individualised study, and what it lacked
A 2019 study at the University of Zurich tailored the target alpha frequency to each patient's own peak frequency rather than using a fixed band, across 15 weekly sessions. Twenty-six people took part and 24 were analysed.
- Distress. Average Tinnitus Handicap Inventory scores fell significantly, from 29.33 before training to 23.92 immediately afterwards, and were 24.75 at six-month follow-up.
- Loudness. Loudness ratings also dropped significantly right after training, from 53.25 to 43.67, but had returned to baseline by six months.
Two things set the size of that. The starting average of 29.33 falls in the mild range of the scale, and the drop was about five points — statistically significant, but smaller than the change patients typically notice, which a 2025 study estimated at about 8 to 12 points (point estimate 11). And there was no control group. The authors name the lack of a placebo control as a limitation of their study, citing time, infrastructure, funding and ethical reasons. Tinnitus distress drifts downward over months in many people without any treatment, so a small before-and-after change with nothing to compare it against cannot be credited to the training.
The randomised trial that tested it
The better test came in 2023, in Brain Communications. Jensen and colleagues randomised 94 people with chronic tinnitus to one of three arms: alpha/delta neurofeedback (31), beta/theta neurofeedback — training a different pair of brainwaves entirely (28) — or a minimal treatment that involved no neurofeedback (35). The neurofeedback groups had 10 sessions over four weeks, with follow-up at three months.
- On distress, measured with the Tinnitus Handicap Inventory, there was no difference between the groups over time. Neither kind of neurofeedback did better than the minimal intervention.
- On intensity, measured with the Tinnitus Magnitude Index, both kinds of neurofeedback reduced rated tinnitus intensity compared with the minimal intervention.
The authors' own summary is that alpha/delta training was sufficient but not necessary for a positive outcome, since training a different pair of brainwaves worked about as well, though they saw a trend towards alpha/delta being more specific. Read beside the Zurich study, the controlled trial reverses its pattern: where the uncontrolled study found lasting distress relief and a fading loudness effect, the randomised one found an intensity effect and no distress advantage over the control. That is what small studies with different designs tend to produce, and it is why neither result should be taken as the answer.
The trial that says it beat CBT
In 2024 Radiology published a randomised trial comparing real-time fMRI neurofeedback — the same principle, but reading brain activity inside an MRI scanner instead of from EEG electrodes — directly against CBT, the best-supported psychological treatment for tinnitus.
The headline is in the paper's title: fMRI neurofeedback outperformed CBT.
The numbers behind it, on the Tinnitus Handicap Inventory:
- At six months — the neurofeedback group fell 28.21 points, the CBT group 12.09 (P = .005)
- At twelve months — the neurofeedback group fell 30 points, the CBT group 4 (P = .01)
The neurofeedback group also improved on sleep and trait anxiety by one month, and on depression and general functioning at six months. The CBT group showed no significant change over time on any of these, with P values running from .14 to above .99.
A 28-point drop is large. Against the thresholds for a change a person notices — about 8 to 12 points on this scale (point estimate 11) — it is not a borderline result. If it holds up, it would be one of the most important findings in tinnitus neuromodulation.
Why it should not yet change what you do
The two arms did not receive comparable treatment. The neurofeedback group had 15 weekly individual sessions. The CBT group had 10 weekly group sessions. That is half again as many appointments, delivered one-to-one instead of in a room with other patients. In an unblinded trial of a self-reported symptom, the amount of individual clinical attention is not a nuisance variable — for several treatments, most of the measured effect is exactly that.
Forty-three people. Twenty-one and twenty-two.
The CBT arm faded. Its 12-point improvement at six months had shrunk to 4 points by twelve, and none of its secondary measures moved significantly. For comparison, the Cochrane review of 28 randomised trials of CBT found it reduced tinnitus impact by the equivalent of 10.91 THI points more than no treatment at the end of treatment, though it had no usable data at twelve months to set the later figure against. When a comparison arm loses most of its gain, the first question is about that arm rather than about the treatment beating it.
And nobody can have this treatment. Fifteen sessions inside an MRI scanner is not a therapy that scales, at any price. Even taken entirely at face value, the practical finding is about mechanism rather than about what to do next week.
The randomised EEG trial above adds a further reason for caution: if training different brainwaves works about as well as training the targeted ones, the active ingredient in neurofeedback may be general rather than specific — and a trial that gave one arm more individual attention than the other is exactly where a general effect would show itself most strongly.
What it deserves is a replication with matched session counts and formats. The trial is registered, randomised and published in a serious journal; it is not to be dismissed. It is to be repeated.
What this means practically
- Neurofeedback requires specialised equipment and a trained provider — it's not something you can meaningfully do without proper EEG sensors and clinical guidance, similar to biofeedback in that regard.
- Do not count on a particular benefit. The uncontrolled study's lasting result was on distress, the randomised trial's was on loudness, and neither was large.
- Protocols ran from 10 sessions over four weeks to 15 weekly sessions, so it is a real commitment of time.
- Try the better-supported options first. CBT, sound therapy and, where there is hearing loss, hearing aids all have more behind them. Neurofeedback is something to ask about if those have not been enough, with modest expectations.