Tinnitus, “the conscious awareness of a tonal or composite noise for which there is no identifiable corresponding external acoustic source,” is a common auditory phenomenon with a complex, multifactorial basis (De Ridder et al., 2021). It is broadly classified as objective or subjective: objective tinnitus, or somatosound, arises from internal bodily sources such as vascular flow (Baguley et al., 2013), whereas this article focuses on chronic subjective tinnitus, the perception of sound without any identifiable internal or external source, which has been present for at least 3 months (Han et al., 2009; Mazurek et al., 2022). Current estimates suggest that more than 740 million adults globally are affected, approximately 14% of the population (Chen et al., 2024). Tinnitus frequently co-occurs with hearing loss and a range of other health problems, and patients vary widely in both symptom severity and hearing profiles (Mazurek et al., 2022). When tinnitus is associated with significant cognitive strain, emotional distress, or disruption of daily functioning, it is classified as tinnitus disorder (De Ridder et al., 2021). In large US samples, roughly one in four individuals with tinnitus report anxiety or depression, and prevalence increases with age, underscoring its particular relevance in aging populations (Tunkel et al., 2014; Bhatt et al., 2017). Evidence from European population studies highlights the central role of comorbidities in shaping tinnitus. For example, non-auditory comorbidities (e.g., metabolic, cardiovascular, and psychiatric disorders) have been linked to tinnitus-related hearing loss and symptom profiles, and these relationships vary systematically with age (Tziridis et al., 2025a). In addition, a UK cross-sectional study identified insomnia, hearing-related distress, and anxiety as the strongest predictors of tinnitus severity, with the highest severity observed in subgroups with the greatest overall comorbidity burden (Beukes et al., 2021). Risk factors associated with tinnitus further include noise exposure, Ménière's disease, modifiable lifestyle factors (e.g., diet), adverse reactions to pharmacological interventions, and head injury or other forms of trauma (Biswas et al., 2023).

The impact of tinnitus extends beyond the individual, imposing substantial burdens on healthcare systems and the economy. For example, a study from the Netherlands estimated nationwide societal costs of €6.8 billion, including €1.9 billion in healthcare expenditure (Maes et al., 2013). A more recent German study estimated annual tinnitus-related socioeconomic costs at approximately €22 billion, a magnitude approaching half of the estimated annual healthcare and productivity costs attributed to diabetes in Germany (approximately €42 billion), highlighting tinnitus as a substantial public health concern (Tziridis et al., 2022).

At present, no treatment has demonstrated consistent, long-term efficacy in reducing the perception of tinnitus. In standard clinical practice, approaches such as cognitive behavioral therapy and conventional sound therapy are mainly designed to improve coping and reduce tinnitus-related distress, with limited emphasis on directly modulating tinnitus-related circuits or inducing targeted neuroplastic change (Han et al., 2021; Mazurek et al., 2022). However, recent advances in mechanistically informed sound-based approaches for selected tinnitus subgroups have begun to show promising benefits (Sendesen and Turkyilmaz, 2024; Yukhnovich et al., 2025; Tziridis et al., 2025b). Overall, the most notable exception remains cochlear implantation, which has been shown to produce sustained reductions in tinnitus loudness, the outcome most sought by patients, but only in individuals with unilateral deafness; given this restricted indication, it lies outside the scope of the present article (Mertens et al., 2016; Husain et al., 2018). To evaluate treatment effects, standardized questionnaires capture different dimensions of tinnitus, including the Tinnitus Handicap Inventory (THI), the Tinnitus Functional Index (TFI), and loudness measured with a Visual Analog Scale (VAS) (Hall et al., 2016; Raj-Koziak et al., 2018). Clinically meaningful improvement is commonly defined as a reduction of at least 13 points on the TFI and at least 7 points on the THI (Zeman et al., 2011; Meikle et al., 2012). Nonetheless, progress in clinical research remains constrained by the absence of established biomarkers and objective outcome measures (Kleinjung et al., 2024).

Reflecting the now widely held view of tinnitus as a disorder of neuroplasticity, recent efforts have therefore shifted toward interventions designed to modulate the underlying neural activity. Neuromodulation applies targeted electrical, magnetic, or mechanical stimuli to shift neuronal excitability, retune aberrant network rhythms, and engage synaptic plasticity that can produce lasting changes in large-scale circuits (Davidson et al., 2024). Among the most widely studied are non-invasive stimulation techniques, such as repetitive transcranial magnetic stimulation (rTMS), various transcranial electrical stimulation (tES) techniques, transcutaneous electrical nerve stimulation (TENS), and neurofeedback approaches. Bimodal protocols that combine sound with vagus nerve stimulation (VNS) or auditory-somatosensory stimulation have also attracted increasing interest. In addition, the emergence of novel technologies in the broader field of neuroscience such as transcranial focused ultrasound (tFUS), transcranial pulse stimulation (TPS), and transcranial photobiomodulation (tPBM) have opened new avenues for tinnitus research that warrant evaluation.

In this article, we review state-of-the-art neuromodulation approaches for tinnitus, outlining mechanisms, clinical evidence, and practical next steps. We then discuss innovative technologies and protocols that expand targets and optimize therapeutic options, with particular attention to multimodal integration and depth-capable targeting.

Tinnitus typically arises from auditory deafferentation following damage to the auditory pathway, most often due to sensorineural hearing loss (Liberman and Kujawa, 2017; Park et al., 2023). It subsequently persists as a systems-level problem and is thought to be driven by an imbalance among three partially overlapping pathways: (i) a lateral auditory pathway that generates and stabilizes the phantom percept's acoustic features; (ii) a medial salience/valuation-attention pathway that determines awareness, priority, and affective load; and (iii) a descending inhibitory pathway that normally gates irrelevant activity before it reaches conscious perception (Elgoyhen et al., 2015; Rauschecker et al., 2010, 2015; Leaver et al., 2016; Krauss et al., 2016; De Ridder and Vanneste, 2021; De Ridder et al., 2022; Knipper et al., 2021; Schilling et al., 2023; see Figures 1A–C). When lateral and medial drive exceed inhibitory control, the tinnitus percept is thought to persist (De Ridder et al., 2022). Moreover, resting-state neuroimaging studies consistently demonstrate aberrant coupling between the auditory cortex and default mode, salience-attention, and executive control networks, supporting a multi-network model of sustained tinnitus perception (Schlee et al., 2008; Carpenter-Thompson et al., 2015; Roberts, 2018; De Ridder et al., 2022; Graetz et al., 2025).

Following deafferentation, neurons in the dorsal cochlear nucleus (DCN) show increased spontaneous firing, enhanced synchrony, and hyperexcitability, changes that propagate rostrally through the inferior colliculus and medial geniculate body (MGB) to the auditory cortex (Shore et al., 2016; Marks et al., 2018; Eggermont and Tass, 2015; De Ridder et al., 2015; see Figure 1A). These alterations have also been interpreted in terms of stochastic resonance, whereby intrinsic neural noise is boosted to compensate for reduced input but can be misread as sound and perceived as tinnitus (Krauss et al., 2016; Schilling et al., 2023). At the auditory cortex, tinnitus is further associated with increased spontaneous activity and abnormal synchrony, while sensorineural hearing loss, often accompanied by tinnitus, induces reorganization of tonotopic maps (van der Loo et al., 2009; Noreña and Eggermont, 2006). At the systems level, resting-state EEG/MEG often shows reduced alpha alongside increased delta-theta and focal gamma over auditory cortices, consistent with thalamocortical dysrhythmia (TCD) models (Llinás et al., 1999; De Ridder et al., 2015; Vanneste et al., 2018). In TCD, reduced sensory input to thalamocortical columns allows alpha activity to slow into theta, while reduced GABA-A mediated lateral inhibition at the border between deafferented and intact frequency maps produces an “edge effect”: low-frequency activity in the deprived zone surrounded by increased gamma, with theta-gamma coupling thought to encode the positive phantom percept (Vanneste et al., 2018). With greater peripheral loss, processing can shift toward parahippocampal areas (auditory memory/context) and thus consolidation (De Ridder and Vanneste, 2021).

Whether the generated auditory pattern becomes consciously intrusive depends on the salience network primarily centered on the anterior insula and dorsal anterior cingulate cortex (dACC) (De Ridder et al., 2011, 2014; De Ridder and Vanneste, 2021; see Figure 1B). This causal involvement is supported by neurosurgical evidence. Early frontal leucotomy studies in tinnitus showed that surgery could markedly reduce tinnitus-related distress without reliably abolishing the tinnitus percept itself (Beard, 1965). More recent work in patients with treatment-resistant depression demonstrates that anterior cingulotomy lesions within the anterior mid-cingulate cortex causally disrupt negative affect processing and cognitive control (Tolomeo et al., 2016). Additionally, conscious auditory perception can be demonstrated only when abnormal activity in the auditory cortex is coupled with engagement of the fronto-parietal attention network (Roberts, 2018; Mazurek et al., 2022). Indeed, tinnitus taxes limited attentional resources and has been associated with reduced selective auditory attention (Roberts et al., 2013; Sedley, 2019; Jensen et al., 2021). Neuroimaging work shows that tinnitus onset is associated with increased theta activity in the pregenual anterior cingulate cortex (pgACC) and reduced theta connectivity between pgACC and auditory cortex, while increased alpha activity in the dACC correlates with tinnitus-related distress (Vanneste et al., 2019, 2024; Chen et al., 2018). A recent machine learning analysis showed that alpha and gamma power together with auditory-limbic connectivity predict susceptibility to brief acoustic suppression, with alpha-band features among the strongest predictors (Shabestari et al., 2025).

Within a descending gating framework, reduced efficacy of pregenual and subgenual medial prefrontal regions (pgACC/sgACC/vmPFC) may fail to recruit the thalamic reticular nucleus to inhibit the medial geniculate relay (“thalamic gating”), permitting aberrant activity to reach and persist in the auditory cortex (Rauschecker et al., 2010, 2015; Leaver et al., 2016; De Ridder and Vanneste, 2021; see Figure 1C). In parallel, the insufficient engagement of brainstem efferent gain control, via midbrain projections (e.g., tectal/periaqueductal regions) that influence the superior olivary complex (SOC), potentially fails to regulate cochlear activity (De Ridder et al., 2012; Terreros and Delano, 2015). Additionally, a paradoxical reward process is thought to contribute to persistence, in which activity within the nucleus accumbens (NAc) coupled with pgACC/vmPFC reinforces tinnitus and biases gating in favor of the ongoing percept (Hullfish et al., 2018; De Ridder and Vanneste, 2021). Preclinical evidence suggests that activating the auditory thalamic reticular nucleus attenuates salicylate-induced hyperactivity in the auditory cortex (Dai et al., 2024). Meanwhile, resting-state fMRI Granger causality showed an association between tinnitus distress and altered bidirectional connectivity between the NAc and the PFC in chronic tinnitus (Xu et al., 2019).

The ascending lateral, medial, and descending pathways form separable yet functionally coupled modules. Bottom-up auditory dysrhythmia in the lateral stream primarily supplies the phantom signal, the medial salience system determines access to awareness, attention capture, and affective load and, in case of failure to inhibit aberrant signals along the descending pathway, the tinnitus persists. Clinically, this three-pathway, systems-level framing motivates circuit-specific, multimodal, and deep-reaching interventions: targeting auditory relays such as the auditory cortex, and the DCN/MGB; recalibrating salience and affective hubs including dACC and pgACC, anterior insula, and parahippocampus; and strengthening the descending gate from vmPFC to the thalamic reticular nucleus and downstream efferents. With this in mind, we shift from “what is disordered” to “what can be rewired,” beginning with a review of established non-invasive neuromodulation approaches.

The shortcomings of prominent unimodal brain stimulation techniques arguably highlight a need to engage broader networks, more diverse and deeper targets along the auditory and frontostriatal pathways, and exploit the mechanisms of neuromodulation-induced neuroplasticity to promote consolidated auditory perceptual learning, sustainably counteracting the tinnitus-related maladaptive changes. Hence, we now turn to approaches designed to engage multiple inputs and deeper nodes of the tinnitus network.

While the exact mechanism underlying bimodal stimulation is still debated, the current consensus suggests a multilevel, timing-sensitive plasticity mechanism: pairing sound with either somatosensory input (e.g., trigeminal/tongue stimulation) or vagal afferent activation drives synaptic plasticity and desynchronization in hyperactive auditory circuits while also engaging diffuse neuromodulatory systems that regulate plastic change (Marks et al., 2018). Thus, bimodal stimulation potentially retunes central gain and synchrony along the auditory pathway, reducing the salience and persistence of the tinnitus percept (Jones et al., 2023).

Foundational preclinical work showed that pairing nucleus basalis stimulation with tones in adult rats can drive pronounced, frequency-specific reorganization of primary auditory cortex maps (Kilgard and Merzenich, 1998). Subsequent work pairing vagus nerve stimulation (VNS) with tones demonstrated similarly targeted plasticity in A1, expanding representation around the paired tone and abolishing behavioral tinnitus proxies, with effects persisting for weeks (Engineer et al., 2011). A prospective randomized, double-blind pilot in implanted patients found greater improvement on the THI with paired VNS than control at 6 weeks (between-group difference not statistically significant) (Tyler et al., 2017). Mechanistically, paired VNS in patients was associated with reduced gamma and increased alpha synchronization in the left auditory cortex, plus decreased theta-band connectivity with dorsal and subgenual ACC and the parahippocampus, with theta-alpha connectivity changes associated with loudness suppression (Vanneste et al., 2017). Meanwhile, pairing transcutaneous VNS with sound has yielded positive, though still preliminary results (Shim et al., 2015; Ylikoski et al., 2020). Clinical outcomes remain variable and require confirmation in larger, sham-controlled trials.

Bimodal auditory-somatosensory stimulation, on the other hand, has gained notable traction over recent years. Pilot work has combined auditory therapy with microcurrent stimulation around the ear (Lee et al., 2024). In a 3-month trial, 60% of patients receiving combined therapy achieved clinically meaningful loudness reductions on VAS compared with 27% in the sound-only group, and two patients in the combined arm reported complete disappearance of tinnitus. Moreover, in a randomized, double-blind trial of 99 patients with somatic tinnitus, bisensory auditory-somatosensory stimulation was delivered via charge-balanced biphasic pulse trains to the skin over the trigeminal ganglion or upper cervical nerves (C1-C2) (Jones et al., 2023). This protocol, designed to induce spike-timing dependent plasticity [long-term depression (LTD)] in cochlear nucleus circuits, produced clinically meaningful improvements in 53%−65% of participants, with significant TFI reductions (−12.0 intent-to-treat; −13.2 per protocol at 6 weeks) and tinnitus loudness that persisted for up to 36 weeks. In contrast, auditory-only stimulation produced no clinically significant effects.

Some of the largest clinical investigations of bimodal auditory-somatosensory stimulation come from the TENT A1, TENT A2, and TENT A3 trials evaluating Neuromod's Lenire device (Boedts et al., 2024). Using charge-balanced biphasic pulse trains delivered to the tongue and time-locked to sound, the TENT A1 trial (n = 326) tested three timing schedules differing in the precision of pairing and the use of background wideband noise (Conlon et al., 2020). All produced significant 12-week improvements (pooled THI −14.2; TFI −13.6), with most gains in the first 6 weeks and a later plateau, suggesting strict millisecond pairing and wideband noise may not be essential. Indeed, TENT A2, a randomized double-blind optimisation study (n = 191), confirmed that pure tones without noise are sufficient, with two bimodal arms showing similar six-week improvements (Arm 1: THI −12.9, TFI −11.6; Arm 2: THI −11.5, TFI −11.7), and showed that resetting parameters at week 6 yielded further benefit by week 12 (pooled THI −18.5; TFI −15.3) that persisted at 12 months (THI −20.2; TFI −17.3); the sound-only arm improved mainly after tongue stimulation was added, supporting a specific somatosensory contribution (Conlon et al., 2022). TENT A3, a multisite trial that ultimately led to Lenire's FDA De Novo approval in 2023 (n = 112), used an active control and found that in moderate to severe tinnitus (THI ≥38) bimodal treatment outperformed sound therapy alone at 6 weeks for THI responders (58.6% vs. 43.2%), while sound alone may suffice initially for milder cases (Boedts et al., 2024). Real-world data from a U.S. cohort (n = 220) reported a 91.5% clinically meaningful improvement at about 12 weeks with a mean THI change of −27.8 and high adherence, though effect size differences likely reflect duration, heterogeneity, and confounds, underscoring the need for cautious interpretation and patient subtyping (McMahan and Lim, 2025). Together, the TENT A1-A3 trials support tongue-sound bimodal stimulation as a clinically useful option for at least a subset of patients (see Figure 3A for an overview).

There is an ongoing debate as to whether precise spike-timing is essential for therapeutic effects in tinnitus. Evidence from Jones et al. (2023) suggests that millisecond-level timing-dependent bisensory stimulation, derived from animal models of LTD in the DCN, produces significant and durable symptom reduction, whereas Conlon et al. (2022) and other studies, including from the broader literature, demonstrate clinical benefits without strict millisecond-level timing (Tyler et al., 2017; Dawson et al., 2021). This suggests that while precise timing may optimize efficacy, co-activations and diffuse neuromodulatory systems possibly also drive clinically meaningful change. Future work should implement adaptive mid-course optimization, stratify and power by baseline severity, pre-specify responder groups, and consider personalized timing and somatosensory targeting to maximize engagement and outcomes. It may also be worth testing additional, broader state-modulating designs that pair transcutaneous stimulation with auditory training to activate neuromodulatory systems and network plasticity beyond strict spike timing rules.

A comparative reading of current neuromodulation modalities suggests a promising future for next-phase tinnitus therapeutics. Open-loop, unimodal cortical stimulation (rTMS and the tES family) has produced reproducible but modest and often short-lived benefits, with outcomes especially sensitive to dose, site, and individual anatomy. Peripheral approaches such as TENS and auricular vagal stimulation are mechanistically plausible and appear helpful in selected phenotypes, yet effects are frequently confounded by expectancy and remain unconsolidated. By contrast, target-aware multimodal strategies, most notably auditory-somatosensory pairing, produce larger and more durable improvements in multiple recent trials, presumably because they induce a “window” of heightened plasticity, which is concurrently leveraged to promote a neural rewiring of auditory circuitry. Moreover, depth-capable methods occupy a complementary niche: traditional options like DBS have yielded large but early-stage effects in refractory cases, while non-invasive ultrasound-based methods (tFUS, TPS) are emerging as practical tools to non-invasively access the MGB, DCN, or frontostriatal hubs with millimeter-scale precision. In parallel, photobiomodulation offers a safe, scalable metabolic and anti-inflammatory intervention that may potentiate paired treatments. Collectively, these early signals justify shifting research from isolated nudges to hypothesis-driven, multimodal, circuit-informed programmes tested in rigorous, placebo-aware trials.

Progress will depend on confronting several persistent challenges head-on. First, heterogeneity remains a major obstacle. Tinnitus is not one disorder, but a family of phenotypes defined by audiometric profile, tinnitus pitch, hyperacusis, somatic modulatability, affective comorbidity, and neurophysiological signatures (e.g., alpha deficits, theta-gamma coupling, and auditory-limbic dysconnectivity). One-size-fits-all protocols arguably dilute effects and obscure responders. Closely linked is the absence of validated biomarkers that stratify patients a priori, confirm target engagement in vivo, and predict durable responses. Additionally, trial design requires optimization. Expectancy and placebo responses are substantial, shams are often inadequate or insufficiently matched, follow-up is commonly short, and parameter reporting is inconsistent. Treatment outcomes need to be consistently reported in a standardized manner, including TFI/THI/VAS scores, and potentially need to be broadened to include neurophysiological endpoints. Finally, the true dose that eventually reaches its (sub)cortical target is rarely verified. Without individualized current flow/field modeling, dosimetry, skull-transmission estimates, or photonic parameter reporting, dose-response relationships will likely remain unclear.

Indeed, the most actionable opportunity to address tinnitus in the near future appears to be principled multimodal integration, an explicitly pragmatic “use all weapons available” stance (De Ridder et al., 2024). In practice, this means combining and sequencing interventions to address ascending lateral, medial, and descending pathways, and to consolidate durable plastic change. The near-term focus should be the optimization of multi/bimodal protocols in light of successful trials with auditory-somatosensory pairing. Protocols should prioritize pairing schedules that proved effective in practice, while also extending parameter settings beyond millisecond spike-timing precision to test alternative, network-level plasticity mechanisms. Promising peripheral targets deserve additional clinical testing, including the greater occipital nerve, which has extensive evidence in pain models and may modulate relevant brainstem nuclei (e.g., nucleus tractus solitarius and locus coeruleus) (Miller et al., 2016; De Ridder and Vanneste, 2017; To et al., 2017). Where home devices permit, it may be valuable to pair or sequence (bimodal) stimulation sessions with neurofeedback blocks that reinforce alpha restoration and reduce excessive theta-gamma coupling, with the aim of consolidating treatment effects. Recently updated safety guidelines on low-intensity electrical stimulation report no additional safety concerns for at-home use (Antal et al., 2025).

Another avenue worth exploring is instructing active rather than passive listening in auditory stimulation. Coupled auditory retraining that requires attention, discrimination, and adaptive difficulty may be more likely to consolidate plasticity than background listening (Hoare et al., 2012; Hemanth and Vipin Ghosh, 2022). Recent sound-based work further suggests that carefully tailored acoustic stimulation can yield measurable benefits in defined tinnitus subgroups. For example, individually matched sound enrichment shaped to the hearing-loss region and tinnitus pitch, and low-intensity noise centered on the tinnitus frequency delivered via hearing aids, have each produced substantial within-group reductions in tinnitus handicap over weeks to months, albeit in relatively narrowly characterized samples (Sendesen and Turkyilmaz, 2024; Tziridis et al., 2025b). Together, these findings indicate that optimizing auditory input (e.g., spectrum, level, dosing, and engagement demands) can produce clinically relevant change in selected phenotypes and should remain a key consideration for the acoustic arm of future bimodal or multimodal neuromodulation protocols.

Given the substantial inter-individual variability, tinnitus trials should also consider exploring closed-loop, brain-state-triggered stimulation. While not yet tested for tinnitus, one study introduced alpha closed-loop auditory stimulation to modulate human alpha power, frequency, and connectivity in a phase-dependent manner (Hebron et al., 2024). Meanwhile, Mulyana et al. (2022) demonstrated that real-time closed-loop tES within a functional MRI framework can optimize alternating-current stimulation frequency and phase between network nodes. Adapting these state-contingent paradigms to gate stimulation on alpha-dominant or reduced theta-gamma states in tinnitus circuits could further enhance target engagement (e.g., in multimodal stimulation contexts). Finally, metabolic priming could also be tested, for example short near-infrared photobiomodulation over A1 just before each session to enhance mitochondrial activity, anti-inflammatory dynamics, and the probability of adaptive plasticity. In parallel, exploring pilots of low-intensity, imaging-guided focused ultrasound to key relays across the lateral auditory and medial salience networks, starting with the MGB and the DCN, and extending to anterior cingulate and ventral striatal hubs, might turn out beneficial. Additionally, such interventions could then be followed by structured bimodal and auditory retraining to consolidate gains.

Methodological upgrades are equally important. Trials should adopt adaptive, multi-arm platform designs that compare sound-based therapy, unimodal stimulation, and multimodal approaches, under a common framework with response-adaptive randomization and one, 6 and/or 12-month follow-ups. Where possible, shams must be sensation-matched and paired with formal blinding checks. Expectancy should be measured and modeled as a covariate rather than treated as negligible. Dose verification should become routine: for tES and rTMS, individual imaging-based current flow and field modeling and reporting of delivered electric field and pulse counts; for tFUS and TPS, ISPTA and ISPPA, duty cycle, focal size, skull correction, and navigation error; for tPBM, wavelength, irradiance, fluence, pulsing, site geometry, and session number/length. A core outcome set should be reported in every trial, including THI/TFI/VAS with minimal clinically important differences, loudness, patient global impression, anxiety, and depression measures, and adherence, potentially complemented by EEG and connectivity endpoints. Finally, an open science posture, including preregistration, harmonized data structures, and routine sharing of raw EEG and analysis code, will accelerate biomarker validation and enable credible meta science.

In sum, the field's recent lessons point in a coherent direction. Tinnitus most likely persists because complex network-level changes maintain a maladaptive percept and its affective load. Therefore, durable relief will seldom follow from a single, open-loop intervention. A strategy that integrates repeated multimodal stimulation, deeper and cleaner targeting, and (closed-loop) personalization, conducted within placebo-aware, dosage-verified, and outcome-rich trials, offers the clearest path to translating mechanistic insight into reliable, generalizable benefit for defined tinnitus subgroups.