For decades, the advice on stopping benzodiazepines was vague. Clinical guidance generally said not to use them for more than about four weeks, but long-term use was, and still is, common.
Patients who had taken the same dose for years often found there was no agreed method for getting off it, and no agreement on what to expect when they tried.
That changed in 2025, and the research published since has started to fill in the picture. Here's what the evidence now says.
Ten medical societies agreed on a number
In 2025, the American Society of Addiction Medicine partnered with nine other medical and professional bodies, including the American Academy of Family Physicians, the American Academy of Neurology, the American Psychiatric Association and the American Geriatrics Society, to publish the Joint Clinical Practice Guideline on Benzodiazepine Tapering. It was funded by a U.S. Food and Drug Administration grant and published in the Journal of General Internal Medicine.
Its central recommendations are unusually concrete:
- Reassess regularly. Clinicians should review the risks and benefits of continued benzodiazepine therapy at least every three months, factoring in age, other conditions, and any concurrent opioid use.
- Never stop abruptly. In patients likely to be physically dependent, sudden discontinuation can cause severe and potentially life-threatening withdrawal.
- Start low and go slow. Begin with a reduction of 5–10% of the total daily dose every 2–4 weeks. The guideline advises not exceeding 25% every two weeks.
- Adjust to the patient, not the schedule. Pace should follow the individual's response. Symptoms after a dose cut are a signal to slow down.
- Sometimes, stop tapering. If withdrawal or other symptoms become intolerable, the guideline explicitly allows for maintaining a patient on a lower, or even the original, dose for an extended period, or indefinitely.
- Add non-drug support. Cognitive behavioural therapy, and CBT for insomnia (CBT-I) in particular, are recommended alongside tapering.
- Most tapers belong in outpatient care. Inpatient or residential management is reserved for severe withdrawal or high-risk cases.
- Older adults are a priority. Tapering should generally be considered for long-term benzodiazepine use in older patients unless there is a compelling reason to continue.
The guideline also states that it is not intended for palliative or end-of-life care.
Two things about that 5–10% figure are worth noting. First, it is considerably slower than the tapers many patients were put through historically. Second, patients with lived experience of benzodiazepine harm sat on the guideline panel and advocated for that slower rate.
A website outperformed usual care by roughly five to one
The guideline says tapering should be a shared decision. A trial published in January 2026 tested what happens when patients are given the information to start that conversation themselves.
The study, published in JAMA Network Open, tested an intervention called EMPOWER-ED, an electronic version of a Canadian patient-education programme. It was run by researchers at Stanford University and the University of Arkansas for Medical Sciences in Veterans Health Administration primary care clinics, enrolling patients who had been on a continuous benzodiazepine prescription for three months or more.
Participants were randomly assigned either to the intervention website or to usual care, which meant continuing to follow their provider's recommendations.
The website explained the risks of long-term use, included first-person accounts from people who had stopped, linked to apps and resources for stress, anxiety and sleep, and generated a customisable tapering schedule. A physician was on call 24 hours a day as a safety net.
Patients who received it were roughly five times more likely to stop benzodiazepines entirely than those in usual care.
This replicates an older finding. The original paper-based EMPOWER trial, published in JAMA Internal Medicine in 2014, found that 27% of older adults who received an educational brochure had discontinued benzodiazepines at six months, compared with 5% of controls, a number needed to treat of four. Sixty-two percent of brochure recipients started a conversation with a doctor or pharmacist about stopping.
The mechanism in both trials is the same, and it is not complicated: most long-term users were never told the risks in a form they could act on, and once they were, many of them raised it themselves.
What happens after the last dose
Acute benzodiazepine withdrawal is well described. What follows it is not.
Some patients report symptoms that persist for months or years after they have fully stopped, anxiety, insomnia, low energy, difficulty concentrating, memory problems, and often report them as new, distinct from whatever the drug was originally prescribed for. In 2023, a working group of 23 experts proposed a name for this: benzodiazepine-induced neurological dysfunction, or BIND.
The strongest evidence to date comes from a scoping review published in PLOS ONE in August 2025. The researchers retrieved 14,097 publications, screened 11,446, and identified 46 studies reporting patient outcomes four or more weeks after complete cessation.
Their conclusion is genuinely two-sided, and worth quoting the shape of:
- Most studies that looked for enduring neurological dysfunction found it.
- Most studies that looked for improvement after discontinuation also found it.
The authors' reading is that stopping benzodiazepines may reveal induced neurological dysfunction in a subset of patients, while conferring long-term benefit on most. Not everyone who takes benzodiazepines develops BIND, and the risk factors are not yet understood.
BIND is now recognised in the 2025 joint tapering guideline, a meaningful shift, because patients describing these symptoms have historically struggled to have them acknowledged.
A note on the evidence: the 2023 survey that introduced the term was a self-selected internet survey of 1,207 benzodiazepine users with no control group, a limitation the authors stated plainly. PLOS ONE published an editorial note on that paper in June 2026. The 2025 scoping review remains the more robust source.
The review also makes a point that often gets lost: benzodiazepines remain genuinely important for specific indications, including catatonia, stiff-person syndrome and certain seizure disorders. "Risky for long-term anxiety management" is not the same as "bad drug."
Does long-term use cause dementia?
This is the question patients ask most, and the honest answer is that the evidence is real but unsettled.
A Canadian case-control study published in the Journal of the Neurological Sciences in 2025, drawing on national health survey data linked to administrative databases, found benzodiazepine use associated with a 65% higher risk of dementia, with long half-life benzodiazepines linked to roughly double the risk of short-acting ones.
The same study identified the central problem. Early dementia symptoms, anxiety, disrupted sleep, are exactly the things benzodiazepines get prescribed for. That's reverse causality, and it inflates any apparent association. The authors found it explained some, but not all, of the excess risk.
Other work points in different directions. A community-based study of older adults found benzodiazepine use associated with mild cognitive impairment but not with dementia. A systematic review published in January 2026 concluded that inconsistent findings across the literature largely reflect differences in study design and unresolved bias rather than a settled effect.
Where that leaves things: a consistent signal for cognitive impairment during use, a plausible but unproven link to dementia, and no randomised trial that could settle it.
The next generation
The underlying pharmacology has been understood for years, and it explains why benzodiazepines are simultaneously so effective and so problematic.
Benzodiazepines are positive allosteric modulators of the GABA-A receptor, they amplify the brain's main inhibitory signal. But they do it non-selectively, hitting several receptor subtypes at once. Broadly: α1-containing receptors drive sedation and amnesia, α2 (and possibly α3) drive the anti-anxiety effect, and α5 receptors, concentrated in the hippocampus, are tied to memory and learning.
Hit only α2 and α3, in theory, and you get the calm without the fog.
The theory has been hard to translate. Alpidem reached market and was withdrawn for liver toxicity. TPA023 was terminated over preclinical toxicity. AZD7325 and PF-06372865 both reached patients and both missed their efficacy targets.
A review published in Medicinal Research Reviews in 2026 argues the tools have now changed. Cryo-electron microscopy has resolved the architecture of individual receptor subunits, and AI-driven modelling has clarified how ligands interact with each one, making it possible to design compounds that separate therapeutic effects from adverse ones far more deliberately than before. The review also highlights newer chemical scaffolds with better metabolic stability and subtype specificity. Among current candidates, KRM-II-81 has shown anxiolytic effects in preclinical work with minimal sedation, respiratory depression or abuse liability.
None of this is available to prescribe. But the target is now specific: keep what benzodiazepines do well, and drop the rest.
The short version
- If you take a benzodiazepine long-term, there is now a real guideline behind stopping, and it is slower and more flexible than what most patients were previously offered.
- 5–10% of the daily dose every 2–4 weeks is the starting point, adjusted to how you respond.
- Never stop abruptly, and don't taper without your prescriber involved.
- Lingering symptoms after stopping are recognised, named, and no longer something you have to argue for.
- Most people who come off benzodiazepines are better for it in the long run.
If tapering is affecting your mental health, tell your prescriber, the guideline treats that as a reason to adjust the plan, not a reason to push through.
Sources
- American Society of Addiction Medicine et al. Joint Clinical Practice Guideline on Benzodiazepine Tapering: Considerations When Risks Outweigh Benefits. Journal of General Internal Medicine, 2025. https://link.springer.com/article/10.1007/s11606-025-09499-2, full guideline: https://www.asam.org/quality-care/clinical-guidelines/benzodiazepine-tapering
- Humphreys K, Hagedorn H, Han X, Kemp L, Poitra N, Cucciare MA. Electronic Intervention for Patient-Managed Benzodiazepine Tapering: A Randomized Clinical Trial. JAMA Network Open. 2026;9(1):e2551807. https://doi.org/10.1001/jamanetworkopen.2025.51807, summary: https://news.uams.edu/2026/02/11/uams-research-shows-online-program-significantly-reduces-benzodiazepine-use-in-adults/
- Tannenbaum C, Martin P, Tamblyn R, Benedetti A, Ahmed S. Reduction of inappropriate benzodiazepine prescriptions among older adults through direct patient education: the EMPOWER cluster randomized trial. JAMA Internal Medicine. 2014;174(6):890–898. https://pubmed.ncbi.nlm.nih.gov/24733354/
- Shade KN, Ritvo AD, Silvernail B, et al. Long-term neurological consequences following benzodiazepine exposure: A scoping review. PLOS ONE. 2025;20(8):e0330277. https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0330277
- Ritvo AD, Foster DE, Huff C, Finlayson AJR, Silvernail B, Martin PR. Long-term consequences of benzodiazepine-induced neurological dysfunction: A survey. PLOS ONE. 2023;18(6):e0285584. (See Editorial Note: PLOS ONE 2026;21(6):e0352305.) https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0285584
- Association between benzodiazepines and dementia: A case-control study from Canadian health surveys and medico-administrative databases. Journal of the Neurological Sciences, 2025. https://www.jns-journal.com/article/S0022-510X(25)00366-1/fulltext
- Unresolved Controversies: The Effect of Benzodiazepines on Cognition and Alzheimer's Disease. Systematic review, published January 4, 2026. https://www.mdpi.com/3042-4518/3/1/2
- Jiang H, Hu Z, et al. The Evolution of Benzodiazepine Allosteric Modulators: Structural Insights From Historical Milestones to Emerging Innovations in Drug Discovery and Development. Medicinal Research Reviews, 2026. https://onlinelibrary.wiley.com/doi/10.1002/med.70026
- Nonsedating anxiolytics (review of subtype-selective GABA-A modulators including TPA023, AZD7325, PF-06372865 and KRM-II-81). ScienceDirect, 2024. https://www.sciencedirect.com/science/article/abs/pii/S0091305724001898
This article is for general information and is not medical advice. Never stop or reduce a benzodiazepine without guidance from your prescriber.