Search for tai chi and fall prevention and you will meet the same sentence over and over: that the US Preventive Services Task Force recommends tai chi for older adults at risk of falling. Open the Task Force’s statement of June 4, 2024, and that sentence is not there. What the panel recommends, at Grade B, is exercise interventions. Tai chi shows up twice on the page, both times describing what some of the underlying trials happened to use, and the Task Force notes that only a smaller number of trials used that kind of three-dimensional movement at all.
The correction costs tai chi nothing, because it has a real endorsement by name. In 2019 the American College of Rheumatology and the Arthritis Foundation strongly recommended tai chi for people with osteoarthritis of the knee, the hip, or both. That put it in the top tier of their guideline, alongside exercise, weight loss, cane use, and anti-inflammatory drugs. Balance exercises and yoga, assessed in the very same document, earned only conditional recommendations. On paper, tai chi outranked them. That ranking no longer holds for yoga. In a summary of updated recommendations approved in September 2026, with the full paper still to come, the American College of Rheumatology conditionally recommends tai chi for knee and hip osteoarthritis, on very low certainty evidence, the same rating it now gives yoga for the knee. The summary does not mention balance exercises.
So the accurate story here is better than the circulating one. What follows is what has been measured, how big those effects are, which comparison group each number came from, and the one experiment that tests whether the practice itself is doing the work or whether any supervised group class would do as well.
The guideline that names it, and the one that does not
A clinical guideline is a committee’s verdict, not a measurement, and the useful parts are the fine print. The rheumatology guideline sorts its recommendations into strong and conditional. Strong means the panel judged the benefits clearly worth the burdens for nearly everybody. Conditional means the evidence was weaker or the trade-off closer, so the choice should be talked through case by case.
In 2019, tai chi landed in the strong column. The panel’s own description of the practice stresses balance, strength, mood and confidence rather than anything mysterious. That same panel declined to grade qigong at all, for two stated reasons: too little evidence, and no agreed definition of the intervention. Article 377 takes up that problem, which is more interesting than it sounds.
Two cautions belong with any citation of this guideline. The panel wrote in its methods that much of the evidence it reviewed was indirect, meaning it did not match the exact question asked, and of low to moderate quality. And the per-recommendation certainty ratings live in tables rather than in the prose. A guideline can issue a strong recommendation on modest evidence, and this one says so about itself.
One other US specialty society names tai chi. The American College of Physicians, in its 2017 guideline on low back pain, lists tai chi among the first things clinicians and patients should try for chronic back pain. Read the parenthesis in that sentence and tai chi sits in the low-quality-evidence bracket, not the moderate-quality one, and the guideline’s literature searches stopped in November 2016.
Federal health agencies are more careful than the secondhand coverage suggests. The Centers for Disease Control and Prevention does include three tai chi programs in its falls compendium, which it presents as a collection of effective community interventions. The page itself, last updated July 28, 2025, contains no sentence declaring that tai chi works. Listing is not endorsing, and the difference matters when a reader is deciding how much weight to give a government logo.
What the falls numbers show
The best synthesis of exercise and falls is a 2019 Cochrane review of 108 randomized trials and 23,407 community-dwelling older adults across 25 countries, average age 76. It reports results by exercise type, and the tai chi row is more interesting than the summaries of it.
Across all exercise types, falls dropped 23 percent, rated high certainty. Reviewers translate that into something you can picture: starting from an illustrative 850 falls per 1,000 people in a year, exercise means about 195 fewer falls. For tai chi specifically, pooled from seven trials and 2,655 participants, the rate of falls fell 19 percent, with a confidence interval running from 0.67 to 0.99. That upper limit of 0.99 just barely excludes no effect, and the reviewers labeled the finding low certainty.
Then the same review reports a second outcome, the number of people who fall at least once, and here tai chi reduced risk by 20 percent at high certainty, from eight trials. Two outcomes, two different confidence levels, from one review. Any article giving you a single tai chi number has dropped half of this.
The comparison the reviewers themselves draw is the one most coverage omits. Plain balance and functional training, the unglamorous stuff physical therapists have prescribed for decades, cut the rate of falls by 24 percent at high certainty, from 39 trials rather than seven. On the second outcome, the number of fallers, the same training cut risk by 13 percent from 37 trials, again at high certainty. Tai chi beats it there: 20 percent against 13, both at high certainty. Cochrane’s conclusion says that programs which reduce falls primarily involve balance and functional exercises, and adds that tai chi may also prevent falls. The order of those two clauses is the finding, and the second outcome complicates it.
A 2020 update by the same lead author, prepared to inform World Health Organization physical activity guidance, added two more tai chi trials and revised tai chi’s own figure to a rate ratio of 0.77 at moderate certainty. That is a genuine disagreement between two documents from the same team two years apart, and it cannot be averaged away. The conservative reading for anyone making a decision is the Cochrane figure with its low-certainty label, noting that the later analysis looked better.
One more result from that review deserves attention precisely because it is negative. Exercise did not measurably improve health-related quality of life. The pooled effect was close to zero and the interval sat well inside what counts as a meaningful difference on standard questionnaires. Falls go down. How people rate their lives does not visibly change.
The experiment that tests the practice itself
Here is the question that tests tai chi: if you took the same people, gave them supervised group exercise at identical frequency, length and duration, and only changed what happened in the room, would the tai chi group still do better?
Somebody ran that experiment. Published in JAMA Internal Medicine in 2018, it enrolled 670 community-dwelling adults aged 70 and over in seven Oregon cities, all of whom had fallen in the previous year or had impaired mobility. Average age was 77.7. Three arms, each two 60-minute classes a week for 24 weeks: a tai-chi-derived balance program, a multimodal exercise program combining balance, aerobics, strength and flexibility, and stretching as the low-dose comparison.
The counts tell the story before the statistics do. Over the trial there were 152 falls in the tai chi arm, 218 in the multimodal exercise arm, and 363 in the stretching arm. Against stretching, both active programs won easily. Against each other, falls were 31 percent lower in the tai chi arm than in the multimodal exercise arm, an incidence rate ratio of 0.69 with a confidence interval of 0.52 to 0.94.
That is the strongest available evidence that something specific to this movement practice, and not merely the fact of showing up to a supervised class, does work on falls.
Two qualifications travel with it and must not be dropped. First, the intervention was not a tai chi class. It was a therapeutically tailored protocol built on the classical concept, using modified forms plus added therapeutic movement exercises, and it is one of the named programs in the federal falls compendium. A reader cannot assume the class at the community center delivers it. Second, the trial’s first author developed that protocol. This is ordinary in behavioral research, the study was federally funded, single-blind and analyzed by intention to treat, and none of that makes the result unsafe to report. A library that discloses conflicts elsewhere should disclose this one.
Where it wins, where it ties, and where the dose is unreal
Pile the good trials together and a pattern appears that is cleaner than either cheerleading or dismissal.
Fibromyalgia, where tai chi beat the standard prescription
A 2018 trial in The BMJ randomized 226 adults with fibromyalgia to supervised aerobic exercise, the usual non-drug recommendation, or to one of four classic Yang-style tai chi schedules. Symptom scores improved in all five groups. Pooled, the tai chi groups improved 5.5 points more on a fibromyalgia impact questionnaire. Matched head to head at the same dose, 24 weeks twice weekly in both arms, the gap was 16.2 points. No serious adverse events related to the interventions were reported.
Knee osteoarthritis, where it tied
A 2016 trial in Annals of Internal Medicine put 204 people with symptomatic knee osteoarthritis into tai chi or a standard course of physical therapy. Both improved substantially. The difference between them was 24 points on the WOMAC scale with a confidence interval from minus 10 to 58, which is to say no difference at all. The tai chi group did better on depression and on the physical side of quality of life. The authors’ word was similar, not superior.
Parkinson disease, where the comparator decides the answer
A 2012 trial in the New England Journal of Medicine randomized 195 patients to tai chi, resistance training or stretching, twice weekly for 24 weeks. On the balance measures tai chi beat both comparators. On falls, it beat stretching and did not beat resistance training. That asymmetry recurs across this whole literature: the margin over doing very little is large, and the margin over genuine exercise is small or absent.
Blood pressure, at a dose almost nobody will do
A 2024 trial in JAMA Network Open randomized 342 adults with prehypertension at two Chinese hospitals to tai chi or aerobic exercise. After 12 months, office systolic pressure was 2.40 mm Hg lower in the tai chi arm. The nighttime ambulatory reading showed a 4.08 mm Hg advantage. Three things belong in the same breath: the between-group difference is small, the dose was four supervised hours every week for a year, and the trial ran entirely in one country whose published results carry a documented pattern discussed in the qigong article.
Breath and fatigue, where the signal is weaker
For chronic lung disease, an overview of 27 systematic reviews covering roughly 37,000 participants, with searches ending in November 2022, singled out exercise endurance as what improved with tai chi in adults with chronic obstructive pulmonary disease. Its authors flagged a structural weakness: most of the reviews they examined discussed results without weighing how biased the original studies were. For fatigue, a 2017 meta-analysis of 10 trials and 689 people found a moderate overall benefit and a smaller one for cancer-related fatigue, but no measurable effect on fatigue from multiple sclerosis or from ageing. Graded moderate quality, with blinding a known weak point.
Nobody can tell you the right amount
Ask how much tai chi you need and the literature cannot answer, which is itself worth knowing.
Across the trials above, weekly supervised time ranges from one hour to four, a fourfold spread. Program length runs from 12 weeks to 12 months. Style is often not even reported: the fibromyalgia trial specifies classic Yang, the Cochrane subgroup pooled seven trials with no style breakdown at all. And the two most striking results in the field, the falls trial and a 2023 cognition trial in Annals of Internal Medicine, both used protocols redesigned for a clinical target. In that cognition trial, standard tai ji quan was the middle arm, and a version with cognitive tasks added beat it by 1.5 points on a 30-point screening test.
The one real dose experiment points somewhere counterintuitive. In the fibromyalgia trial, 24 weeks of tai chi beat 12 weeks by 9.6 points, while doing it twice a week instead of once added nothing measurable. Months mattered. Sessions per week did not.
In 2025 the investigators who lead this field published a two-part white paper drawn from a conference held in late 2023 at Harvard Medical School’s Osher Center for Integrative Health. Their own list of unsolved problems begins with styles, training regimens and dosage, then adds contextual effects, cost-effectiveness and teacher credentialing. That is the people who study tai chi saying the manual has not been written yet.
Qi is the tradition’s account, not the trial result
Tai chi comes out of Chinese tradition, which describes the practice as cultivating and circulating qi, a vital energy understood to move along channels in the body. That is what the tradition holds, in the tradition’s own vocabulary, and reporting it is a matter of accuracy.
It is also a separate question from anything above. A lower fall count in an Oregon trial is evidence about falls. It is not evidence about qi, and nobody has designed a trial that could be. The rheumatology panel modeled the right move: it described the tradition in one sentence, then explained the measured benefit in terms of strength, balance, mood and self-confidence.
Read the direction of error in both directions. Treating measured benefits as proof of an energy system overclaims. Treating an unproven energy system as grounds to ignore a federally funded randomized trial underclaims. Both are the same mistake about what a trial can show.
Safety, and why the clean record is half a sentence
Every randomized trial described above reports the same result on harms: no serious adverse event attributable to the intervention. That is consistent across fibromyalgia, knee osteoarthritis, Parkinson disease and falls, in trials run by different teams in different cities. The National Center for Complementary and Integrative Health, whose tai chi page was last updated in December 2023, says the practice appears to be safe and that reported problems were minor, mostly aches in muscles and joints.
Now the other half. The Cochrane reviewers found that adverse events were reported to any degree in 27 of 108 falls trials, and closely monitored in both groups in exactly one. Fourteen trials reported none. The authors of a 2017 meta-analysis of tai chi for fatigue were blunter: the safety of tai chi cannot be evaluated because the data are absent. A clean record from a literature that mostly was not looking is encouraging rather than conclusive.
The practical risk is who is in the room and what else is going on. Get a clinician’s input first if you have had a fall you did not understand, if standing up makes you dizzy, if you have a joint replacement or an unstable fracture site, if a neurological condition is changing month to month, or if chest discomfort or breathlessness shows up during mild exertion. New dizziness, a fall during class, or pain that outlasts a session by days are reasons to stop and call, not to push through.
Five questions worth asking before the first class
There is no licence to check. Tai chi teachers are not regulated by any state or by the federal government, and no national certification carries authority. Various organizations train and certify, using criteria they each set. That means the checking is yours to do, and these five questions get at what the trials controlled.
- How long is the course, and does it run for months rather than weeks? Duration is the one dose variable with real evidence behind it.
- Is anyone supervising, and how big is the class? Every trial here used supervised instruction, most in small groups.
- Does the teacher adapt forms for people with knee replacements, poor balance or Parkinson disease, and have they done it before?
- Is a chair available and is standing optional? Trial populations included people with impaired mobility.
- What happens if you fall or feel unwell in class, and will the teacher talk to your physical therapist or physician?
If a teacher answers the last question with a promise to treat a diagnosed condition, that is a different conversation. Nothing here is a reason to stop, delay or substitute any treatment, and that decision belongs with your clinician.
The one thing to take away
If you want the shortest honest summary: tai chi is a legitimate, low-risk form of exercise with real, guideline-recognized benefit for osteoarthritis of the knee and hip (a conditional recommendation since 2026), measurable benefit for falls and balance, and a win over the standard non-drug prescription in fibromyalgia. It reliably beats stretching and doing nothing. Against real exercise and real physical therapy it sometimes wins and often ties. Pick it because you will keep doing it for six months, not because it outranks the alternatives, and bring whoever manages your health into the decision before you start.
Sources
- US Preventive Services Task Force. “Falls Prevention in Community-Dwelling Older Adults: Interventions. Final Recommendation Statement, June 4, 2024 (Grade B for exercise interventions; tai chi appears only in descriptions of trial content).” uspreventiveservicestaskforce.org
- Arthritis & Rheumatology (American College of Rheumatology / Arthritis Foundation). “2019 American College of Rheumatology/Arthritis Foundation Guideline for the Management of Osteoarthritis of the Hand, Hip, and Knee (tai chi strongly recommended for knee and/or hip OA; balance exercises and yoga conditional). 2020. DOI: 10.1002/art.41142 (updated by the 2026 ACR recommendations; cited for its 2019 positions).” pubmed.ncbi.nlm.nih.gov
- American College of Rheumatology. “2026 Update of the American College of Rheumatology Recommendations for the Management of Osteoarthritis of the Knee, Hip, and Hand (guideline summary approved by the ACR Board of Directors September 8, 2026; full manuscript to be submitted; tai chi conditional for knee and hip OA, very low certainty; yoga conditional for knee OA, very low certainty).” assets.contentstack.io
- Annals of Internal Medicine (American College of Physicians). “Qaseem A, Wilt TJ, McLean RM, Forciea MA. Noninvasive Treatments for Acute, Subacute, and Chronic Low Back Pain: A Clinical Practice Guideline From the American College of Physicians (tai chi named in the low-quality-evidence bracket). 2017. DOI: 10.7326/M16-2367.” pubmed.ncbi.nlm.nih.gov
- Centers for Disease Control and Prevention. “Falls Compendium, National Center for Injury Prevention and Control (page last updated July 28, 2025; three tai chi programs listed, no effectiveness claim in prose).” cdc.gov
- National Center for Complementary and Integrative Health (NIH). “Tai Chi: What You Need To Know (content last updated December 2023).” nccih.nih.gov
- Cochrane Database of Systematic Reviews. “Sherrington C, Fairhall NJ, Wallbank GK, et al. Exercise for preventing falls in older people living in the community (108 trials, 23,407 participants; tai chi subgroup). 2019. DOI: 10.1002/14651858.CD012424.pub2.” pubmed.ncbi.nlm.nih.gov
- International Journal of Behavioral Nutrition and Physical Activity. “Sherrington C, Fairhall N, Kwok W, et al. Evidence on physical activity and falls prevention for people aged 65+ years: systematic review to inform the WHO guidelines. 2020. DOI: 10.1186/s12966-020-01041-3.” pubmed.ncbi.nlm.nih.gov
- JAMA Internal Medicine. “Li F, Harmer P, Fitzgerald K, et al. Effectiveness of a Therapeutic Tai Ji Quan Intervention vs a Multimodal Exercise Intervention to Prevent Falls Among Older Adults at High Risk of Falling. 2018. DOI: 10.1001/jamainternmed.2018.3915.” pubmed.ncbi.nlm.nih.gov
- The BMJ. “Wang C, Schmid CH, Fielding RA, et al. Effect of tai chi versus aerobic exercise for fibromyalgia: comparative effectiveness randomized controlled trial. 2018. DOI: 10.1136/bmj.k851.” pubmed.ncbi.nlm.nih.gov
- Annals of Internal Medicine. “Wang C, Schmid CH, Iversen MD, et al. Comparative Effectiveness of Tai Chi Versus Physical Therapy for Knee Osteoarthritis: A Randomized Trial. 2016. DOI: 10.7326/M15-2143.” pubmed.ncbi.nlm.nih.gov
- The New England Journal of Medicine. “Li F, Harmer P, Fitzgerald K, et al. Tai chi and postural stability in patients with Parkinson's disease. 2012. DOI: 10.1056/NEJMoa1107911.” pubmed.ncbi.nlm.nih.gov
- JAMA Network Open. “Li X, Chang P, Wu M, et al. Effect of Tai Chi vs Aerobic Exercise on Blood Pressure in Patients With Prehypertension: A Randomized Clinical Trial. 2024. DOI: 10.1001/jamanetworkopen.2023.54937.” pubmed.ncbi.nlm.nih.gov
- Annals of Internal Medicine. “Li F, Harmer P, Eckstrom E, Fitzgerald K, Winters-Stone K. Clinical Effectiveness of Cognitively Enhanced Tai Ji Quan Training on Global Cognition and Dual-Task Performance During Walking in Older Adults. 2023. DOI: 10.7326/M23-1603.” pubmed.ncbi.nlm.nih.gov
- PLOS One. “Xiang Y, Lu L, Chen X, Wen Z. Does Tai Chi relieve fatigue? A systematic review and meta-analysis of randomized controlled trials. 2017. DOI: 10.1371/journal.pone.0174872.” pubmed.ncbi.nlm.nih.gov
- Disability and Rehabilitation. “Heredia-Rizo AM, Martinez-Calderon J, Pina-Pozo F, Gonzalez-Garcia P, Garcia-Munoz C. Effectiveness of mind-body exercises in chronic respiratory diseases: an overview of systematic reviews with meta-analyses. DOI: 10.1080/09638288.2023.2233899.” pubmed.ncbi.nlm.nih.gov
- Journal of Integrative and Complementary Medicine (Harvard Medical School Osher Center for Integrative Health). “Yeh GY, Ahn A, Clark J, et al. The Science of Tai Chi and Qigong as Whole Person Health, Part II: Evidence Gaps and Opportunities for Future Research and Implementation. 2025. DOI: 10.1089/jicm.2024.0958.” pubmed.ncbi.nlm.nih.gov