Hypnotherapy for tinnitus looks effective only against weak comparisons: it did well against a waiting list, but a 1996 randomised trial of 86 patients found it no better than a single counselling session on any quantitative measure.
It has a longer research history than most alternative treatments — trials date back to the 1980s. What that history shows depends almost entirely on what each trial used as its comparison, and the pattern is clean enough to be worth setting out directly.
Against nothing much, it looks good
A 1990 study compared self-hypnosis against two controls: a brief auditory stimulus presented to the tinnitus ear, and a waiting list. 73% of the self-hypnosis group reported their tinnitus disappearing during treatment sessions, against 24% of the brief-stimulus group, and only the self-hypnosis group showed significant improvement in symptom profiles at one week and two months.
A 1993 study from the same group took 45 male patients with tinnitus from acoustic trauma and assigned them to self-hypnosis, masking, or attentiveness to the patient's complaints. Self-hypnosis significantly reduced tinnitus severity. Masking did nothing significant. And attentiveness — a clinician listening — partially relieved the tinnitus.
That third arm is the interesting one, and it is a warning about the first study rather than a footnote to the second.
Against an ordinary conversation, it looks like nothing
In 1996 a randomised trial in The Journal of Laryngology and Otology asked the question directly: is hypnotherapy better than counselling? Forty-four patients received three sessions of client-centred hypnotherapy; 42 received a single counselling session.
Outcomes were tinnitus loudness match, a subjective symptom severity score, the trend of a linear analogue scale, requests for further therapy, and whether patients felt improved.
On all three quantitative measures, and on requests for further follow-up, client-centred hypnotherapy was no better than counselling. The one difference was the subjective one: 45.5% of the hypnotherapy group reported a general sense of improvement, against six patients in the counselling group.
The authors' own reading of that gap is the sentence worth carrying away — their study, they wrote, could not demonstrate whether it was a genuine hypnotic effect or simply a response to the additional attention from the therapist. Three appointments against one buys more attention regardless of what happens inside them.
This is the comparison problem in its clearest form. A waiting list controls for time passing; a brief tone controls for sitting in a clinic. Neither controls for a person paying attention to you, which the 1993 study had already shown does something on its own.
The smallest trial is the oldest, and it is instructive
A 1985 controlled trial took 14 patients with constant unilateral tinnitus that had resisted every other therapy, and gave each of them hypnosis in three forms in random order — with trance induction alone as the control arm, compared against "ego boosting" and active suppression of the tinnitus under trance.
One of the 14 responded significantly to active suppression. Five of the 14, 36%, found the induction of a hypnotic state useful — it helped them tolerate the tinnitus, while its loudness and quality were unaltered.
Fourteen patients cannot settle anything. But the shape of the result — a third of people coping better, almost nobody changing the sound, and the control arm doing much of the work — has been reproduced by everything since.
Where the striking numbers come from
The most quoted figures for hypnotherapy come from uncontrolled work. A non-randomised clinical trial followed 35 patients through Ericksonian hypnosis, five to ten sessions teaching self-hypnosis, and reported average Tinnitus Handicap Inventory scores falling from 60.2 to 16.9 by discharge, across a group including patients rated severe and catastrophic at baseline.
A 43-point drop is enormous — several times the amount a person needs to notice a change. It is also a before-and-after measurement with no control arm at all, in a condition that improves on its own and in a treatment whose main measurable ingredient may be clinical attention. The number is real; what it measures is not established.
What the meta-analysis says about the family
A 1999 meta-analysis pooled 18 studies and up to 700 people across cognitive and cognitive-behavioural treatment, relaxation, hypnosis, biofeedback, education and problem-solving.
Psychological treatment works on tinnitus annoyance: d = 0.86 in the controlled studies, falling to 0.5 in before-and-after designs and 0.48 at follow-up. Effects on tinnitus loudness were weaker and disappeared at follow-up entirely — the recurring finding on this site that these treatments change how tinnitus affects you rather than how loud it is.
And the part that explains the guidelines: in exploratory analysis, cognitive-behavioural treatments were more effective on annoyance than the alternatives in the controlled studies. Hypnotherapy is inside a family that works, and it is not the part of the family that worked best.
What this means practically
It is not dangerous, and it is not fringe. It has been studied in randomised trials, which is more than several treatments marketed alongside it can claim.
But the evidence does not support choosing it over the alternatives. CBT, sound therapy, and the approaches in choosing between them are guideline-supported and better evidenced, and cognitive-behavioural treatments outperformed the rest of the psychological family in the one analysis that compared them. If what draws you to hypnotherapy is a calmer, less effortful kind of therapy, mindfulness and acceptance-based therapy has randomised trials against active comparisons behind it.
If you try it anyway, the expectation the evidence supports is that you may well feel better afterwards, that this is a real and worthwhile thing to feel, and that the 1996 trial suggests a single honest conversation about your tinnitus might have done the same.
A note on what this page used to say
An earlier version described hypnotherapy as "promising but under-evaluated by the standards used to write clinical guidelines", and said explicitly that this was different from evidence showing something does not work. That was wrong about the literature: a randomised trial of 86 patients had compared hypnotherapy against counselling and found no advantage on any measured outcome. The page also quoted the 73%-against-24% figure without citing the study it came from. Both are fixed, and the corrections log records it.