Guided internet CBT reduces tinnitus distress: in a randomised trial it beat weekly monitoring, and pooled across nine studies it improved the Tinnitus Functional Index by 12.48 points, though not the Tinnitus Handicap Inventory.
Of everything on the treatment comparison, cognitive behavioural therapy has the strongest evidence. The Cochrane review says so. The VA/DoD guideline recommends it. And most people with distressing tinnitus never get offered it.
That gap — best-supported and least available — is the entire reason internet-delivered CBT matters.
The problem, stated by the researchers
The trial's own framing is blunt: managing tinnitus is notoriously difficult, access to evidence-based care is limited, CBT is the tinnitus management strategy with the most evidence of effectiveness, and it is rarely offered to those distressed by tinnitus. Internet delivery overcomes the accessibility barrier; at the time the study ran, it was not readily available in the United States.
None of that is a claim about efficacy. It is a claim about supply, and it is the thing most treatment articles leave out.
What the trial did
- Two arms, randomised 1:1: audiologist-guided internet-based CBT (n = 79) against a weekly monitoring group (n = 79).
- Recruited online, through an open-access website, in English or Spanish.
- Primary outcome: change in tinnitus distress on the Tinnitus Functional Index.
- Secondary outcomes: anxiety, depression, insomnia, tinnitus cognitions, hearing-related difficulties, quality of life.
- Two-month follow-up to test whether the effect held.
Blinding was not possible, which the authors state. That is inherent to the intervention — you cannot conceal from someone whether they are doing a therapy programme.
What it found, and how to size it
Internet CBT led to a greater reduction in tinnitus distress: a mean TFI of 36.57 (SD 22) against 46.31 (SD 20.63) in the weekly monitoring group.
About ten points apart on the instrument the field uses. Here is how to read that honestly:
- It is a between-group difference at endpoint, not an individual's change score. The 13-point threshold the TFI's developers set for meaningful improvement applies to how much one person's score moves, so the two numbers are not directly comparable. How tinnitus is measured sets out why that distinction matters.
- The comparator was weekly monitoring, not nothing and not face-to-face CBT. Monitoring is a real control — people were being asked about their tinnitus every week, which is itself an intervention of sorts — but it is a lower bar than a clinic.
- The secondary outcomes moved too, across anxiety, depression, insomnia and tinnitus cognitions, and the effects were stable at two months. A treatment that shifts one number and nothing else is a weaker result than one that shifts the cluster.
So: real, broad, durable at two months, and moderate in size. That is a fair summary and it is a better result than most things on this site get.
Nine trials pooled, and two outcomes that refuse to move
A 2025 systematic review and meta-analysis in Explore gathered nine studies of internet- and mobile-device-delivered CBT for tinnitus — the wider view this article previously lacked. It is more interesting than a straight confirmation, because the outcomes split in two.
What improved significantly:
- Tinnitus Functional Index — mean difference −12.48 (95% CI −16.43 to −8.54)
- Tinnitus Questionnaire — −5.52 (−8.15 to −2.88)
- Insomnia Severity Index — −2.65 (−3.64 to −1.67)
- HADS anxiety — −1.92 (−2.72 to −1.12), and HADS depression — −1.40 (−2.18 to −0.61)
- GAD-7 — −1.33 (−2.37 to −0.30), and PHQ — −1.46 (−2.87 to −0.06)
What did not:
- Tinnitus Handicap Inventory — −2.98, 95% CI −12.03 to 6.07
- Tinnitus cognitions questionnaire — −9.09, 95% CI −23.46 to 5.28
Why the two main tinnitus questionnaires disagree
The TFI moved by 12.48 points. The THI moved by 2.98, with an interval running from −12 to +6 — a range that includes getting somewhat worse.
Both are standard measures of tinnitus severity, applied to the same treatment. The measurement article linked above explains how they can come apart: the THI was published in 1996 to assess handicap at a point in time, and the TFI was built in 2012 precisely because the field lacked an instrument validated to detect change. A treatment that moves the responsive instrument and not the older one is the pattern that difference predicts.
That is an explanation rather than a dismissal, and it cuts both ways. It means the THI null is weaker evidence against i-CBT than it looks — the interval is enormous, which usually signals that few of the nine studies reported it. It also means the positive result rests substantially on the instrument chosen to measure it, which is worth knowing when a single number gets quoted at you.
One mild tension with the trial described above: Beukes et al. reported improvement in tinnitus cognitions, and the pooled analysis across nine studies does not confirm it. That is the ordinary fate of secondary outcomes as more studies are added, and a reason to weight the pooled picture over any single trial.
The size of it, held against the threshold
The pooled TFI difference of 12.48 points invites comparison with a threshold — and which threshold you pick changes the answer, which is worth being straight about.
Against the TFI developers' preliminary 13-point criterion, 12.48 falls just short. Against the 9-point estimate a 2025 multi-centre study derived by anchoring to patients' own ratings of improvement, it comfortably clears. The same measurement article sets out both and why they differ.
On the better of the two estimates, this result is above the threshold rather than below it — which is a more favourable reading of internet CBT than this page previously gave, and the correction belongs here rather than buried.
That comparison is not quite legitimate, for the reason set out earlier on this page: a pooled between-group difference and one person's change score are different quantities. But it is worth registering how consistently tinnitus treatments land near that line rather than far above it — and this is one of the better results in the field.
The mood and sleep numbers deserve the same calibration. A 1.33-point drop on the GAD-7, which runs to 21, and 1.40 on the HADS depression subscale, which runs to 21, are real, statistically solid and small. What the pooled evidence supports is a broad, modest improvement across distress, sleep and mood — not a transformation of any one of them.
The wider digital picture, and the finding that should change how you read it
Three more reviews published in 2025 and 2026 widen this out beyond CBT specifically, and one of them contains the most useful single observation on this page.
The pooled effect. A 2025 systematic review and meta-analysis in Telemedicine and e-Health covering e-health and internet-based psychological interventions found a large effect on tinnitus distress (d = 0.83, 95% CI 0.61–1.06) across 450 people in experimental groups against 504 controls, and a moderate effect on tinnitus handicap (d = 0.59, 0.44–0.73). Depression, anxiety and insomnia improved less strongly but significantly, and most long-term outcomes remained stable.
The breadth. A 2026 update of a telerehabilitation review in the Journal of Medical Internet Research gathered 24 papers across guided internet CBT (5 studies, 619 people), self-help manuals, self-help devices, smartphone apps (13 studies, 23,788 people) and other internet interventions. All categories came out effective at reducing tinnitus severity and distress on questionnaires.
And now the finding that reframes both. The meta-analysis tested what predicted the size of the effect, and risk of bias was one of the two moderators it identified — alongside tinnitus severity.
Sit with that. It means the studies at higher risk of bias produced the larger effects, measurably, within this literature. Not as a suspicion or a general principle, but as a variable the analysis could fit. The d = 0.83 headline is inflated by exactly the studies least able to support it, and the honest estimate lies somewhere below it.
The telerehabilitation review reaches the same place from its own direction: every included study raised some to great concerns of risk of bias, its overall certainty of evidence ran low to moderate, and not all studies accounted for confounders as basic as hearing loss, hyperacusis, anxiety, depression or sleep problems.
Where these reviews disagree, and why it is worth knowing
The e-health meta-analysis concludes that internet-based therapy is valuable as an initial therapeutic contact and a supporting component when accompanied by therapists — and explicitly names the heterogeneous quality, high dropout and risk of bias as reasons not to treat it as first-line management.
That sits slightly against the Tonndorf Lecture summarised on the CBT page, which describes unguided internet CBT as effective with dropout as its main drawback.
They are not irreconcilable — one is describing what a well-run programme can achieve and the other what the literature as a whole will support — but the difference is real, and the cautious reading is the better one to act on: digital delivery is well evidenced as a route into care and thinner as a replacement for it.
Why people stop, which is the actual limit
Both reviews name dropout without explaining it. A 2025 mixed-method systematic review in Translational Behavioral Medicine looked specifically at engagement with digital health interventions for adults with hearing loss and tinnitus, and found what determines it is largely user-side: accessibility, whether the intervention empowers the person using it, and whether it fits their actual needs and lifestyle.
Its own caveat is honest — few studies treat engagement as a primary outcome, so its inferences rest on secondary findings.
That is still the practically useful frame. The evidence above describes what these programmes achieve in people who complete them. Whether you complete one depends less on the evidence behind it than on whether it fits into your life, which is worth weighing before starting rather than discovering at week three.
What the VA/DoD guideline says about online CBT
The 2024 VA/DoD guideline does not give internet CBT a recommendation of its own. It folds it into recommendation 14, which suggests CBT by a trained provider for adults with bothersome tinnitus (weak for), and its reasoning is worth knowing.
- It drew on a systematic review of 10 randomised trials with 1,194 people that looked at CBT delivered other than in person: nine trials of internet CBT, one by telephone, and one self-administered from books. These formats might reduce tinnitus distress compared with passive controls.
- It found that CBT delivered remotely, CBT in person, and self-administered CBT with some provider involvement led to similar reductions in tinnitus distress, because no differences had emerged in the head-to-head comparisons available.
- In one trial it describes, people randomised to in-person group CBT were more satisfied and rated it more helpful than those given internet CBT, though both groups improved similarly against a discussion-forum control and kept their gains at six months.
- Separately, recommendation 4 finds insufficient evidence to recommend for or against web-based or app-based self-management — the unguided category, as distinct from CBT with a provider involved.
The guideline rated its confidence in the CBT evidence as very low, citing small samples and imprecise effect sizes, while noting that CBT has more studies and more consistent findings than other psychological treatments for tinnitus. It also names the reason delivery format matters so much: behavioural and mental health providers who can deliver CBT for tinnitus are in limited supply in the United States.
The detail that matters practically
It was guided by an audiologist.
Not a clinical psychologist. That single design choice is what could make this scale, because the bottleneck for CBT in tinnitus has never been the evidence — it is that psychologists trained in tinnitus are rare and audiologists are who people with tinnitus already see.
The same research group has run a Spanish-language feasibility trial and a US pilot, which is what building towards availability looks like rather than a single study.
What this does not say
It does not say an unguided app is equivalent. The thing tested was guided internet CBT, with a clinician in the loop. Tinnitus apps and self-directed programmes are a different category and are not what this evidence supports.
On its own, it does not say online CBT matches in-person CBT; that was not the comparison it made. The guideline evidence above is the closer answer, and it points to similar results by either route.
And it does not say the sound gets quieter. As with all CBT for tinnitus, what improves is distress, intrusion and the surrounding load — which is what the instruments measure and what changes how people live.
What to do with this
- Ask an audiology service whether they offer or can refer to a guided online CBT programme for tinnitus. That is the version with the trial behind it.
- If you are offered face-to-face CBT, take it — the online route exists because the in-person route is scarce, not because it is better.
- Be wary of unguided apps marketed as CBT — while keeping the distinction between an app and a programme. The guidance was part of the intervention tested in the trial above, and a 461-patient international trial has since tested app-based counselling and app-based sound therapy delivered with no clinician contact, finding them the weaker components among its arms. But unguided internet-based CBT — a structured therapeutic programme completed without a clinician — is itself described as effective in the 2025 Tonndorf Lecture, with higher dropout rates as its main drawback. And one specific app now has a sham-controlled trial behind it, covered under tinnitus apps. The line worth drawing is not guided against unguided; it is a programme that has been tested against a product that borrows the name.
- Expect the shape of the benefit to be less distress rather than less sound, and judge it on that.