Does brain training work for older adults? What the evidence shows in 2026

Why cognitive aging is now a prevention question

World Health Organization figures cited in a 2026 study put dementia at approximately 55 million people globally, with nearly 10 million new cases a year; the over-60 population is projected to more than double between 2020 and 2050, from 1 billion to 2.1 billion^[3]. A 2024 scoping review projects 14,298,671 people affected in the European Union alone by 2050^[1].

Mild cognitive impairment (MCI) affects an estimated 15% to 23% of community-dwelling adults aged 50 and over, and predicts progression to dementia in up to 46% of cases within 3 years^[3]. Yet at least 40% of dementia cases are associated with modifiable causes^[3], which is why researchers describe a "window of opportunity for delaying the development of dementia, especially considering the absence of a known cure"^[1]. Cognitive training is one candidate strategy under study^[1].

Does computerized training improve cognition after 65?

The cleanest test is a four-condition randomized controlled trial in Frontiers in Aging Neuroscience^[2]. Healthy volunteers at a retirement community in Lenexa, Kansas, 59% of them aged 80 or older (range 65–93), were randomized to 4 months of computerized cognitive training (n = 33, mean age 80), mild aerobic training (n = 29, mean age 81), both combined (n = 29, mean age 79), or a book-reading club (n = 31, mean age 79). The cognitive arm trained 32 hours in total: 48 forty-minute sessions, three per week for 16 weeks.

The result was one-sided. Cognitive-training participants, alone or combined with exercise, improved significantly on Hand-Eye Coordination, Global Visual Memory (working memory and long-term memory), Speed of Information Processing, Visual Scanning, and Naming, with Cohen's d of 0.6–0.7 (medium) to 0.8 (large)^[2]. Neither the aerobic-only nor the book-reading group improved on any of the 13 abilities measured^[2].

Of 180 randomized participants, 122 (67.8%) completed; the cognitive-training arm had the lowest dropout of the four groups at 26.7%, and non-completers were on average 4 years older^[2].

What do the newest studies (2022–2026) add?

A 2026 quasi-experimental controlled study from the Universidad de Castilla-La Mancha in Spain followed 45 community-dwelling adults over 60 (23 intervention, mean age 70.5; 22 control, mean age 71.5) through 12 weeks of app-based training at home, 36 sessions of 20–30 minutes, three per week^[3]. The trained group gained 132 points (SD 98.55) on a 0–800 global cognitive score, versus 18 points (SD 71.53) in controls (p < .001), with significant gains also on the Montreal Cognitive Assessment^[3]. Everyday memory failures did not improve, but the trained group avoided the deterioration seen in controls^[3]. Of 60 participants who completed the initial evaluation, 45 (75%) returned for final assessment 8 weeks after training ended^[3].

Adults who already have cognitive impairment were tested in a single-blind randomized study in BMC Geriatrics^[4]. Forty-seven adults (mean age 81.2, SD 8.2) attending adult day centers were randomized to 8 weeks of thrice-weekly 40-minute sessions of either a movement-based group program ("Thinking in Motion", n = 28) or individual computerized training (n = 19). Global cognition improved significantly from baseline to post-intervention, main effect of time F(1,44) = 17.43, p < .001, partial η² = .283, with no difference between formats (p = .970)^[4]. The conclusion is blunt: both trainings "improve cognition but not gait in older adults with cognitive impairment"^[4].

Is brain training plus exercise better than either alone?

The four-arm 2013 trial was designed to answer this, and the answer was no, at least for mild aerobic activity. The combined group received twice the training time of either single intervention, yet its improvements (Eye-Hand Coordination, Naming, Speed of Visual-Spatial Processing) were a subset of what the cognitive-only group achieved, and the absence of significant three-way interactions suggested "the few improvements observed in the combined intervention group are driven solely by the cognitive training"^[2].

The 2022 randomized study points the same way: the co-dependent motor-cognitive program was as effective as screen-based training for cognition, but neither moved gait speed or stride-time variability under single- or dual-task conditions^[4].

How much training do the studies actually use?

A 2024 PRISMA scoping review in The Journal of Prevention of Alzheimer's Disease screened 1,733 records and included 34 studies of smartphone and web-based cognitive training in older adults, spanning 27 apps^[1]. Doses ranged from 20 to 600 minutes per week, sessions from 10 to 240 minutes, frequency from twice weekly to daily, and duration from one intensive week to 12 months, two sessions per week being the most common^[1]. The trials above sit mid-range: 60–120 minutes per week over 8–16 weeks^[2][3][4].

What gets trained is more consistent: 79.4% of the 34 studies targeted memory, 76.5% executive functions, and 73.5% attention; language (35.3%), praxis (26.5%), visuospatial abilities (11.8%) and orientation (5.9%) trail far behind^[1]. The reviewers conclude that "further research is essential to determine which cognitive training interventions work best for different populations and cognitive domains"^[1].

What this evidence does not show

Four caveats. First, samples are small: the scoping review found a mean of roughly 39 participants per study, with only 4 of 34 studies enrolling 100 or more^[1]. Second, the 2026 Spanish study allocated participants non-randomly, and its authors list the nonrandomized design, small sample, and self-reported measures as limits on internal validity^[3]. Third, authorship matters: the 2013 trial was authored by a researcher affiliated with the company that donated the training program^[2].

Fourth, and most important: none of these studies shows that brain training prevents dementia. They show improvement on trained cognitive abilities and standardized assessments over 8 weeks to 4 months. Transfer beyond cognition was absent where measured, no gait change in the impaired cohort^[4], no significant movement in depression, sleep, or quality of life in the Spanish cohort^[3]. Protocol variation of 20 to 600 minutes per week "may lead to inconsistent findings and limit the ability to compare and generalize the results"^[1]. Longer, larger randomized trials with follow-up are the missing piece.