Are brain training apps effective? Mostly at one thing: making you better at the games inside the app. The large reviews agree that practice raises scores on the trained tasks, sometimes on closely similar tasks, and rarely or never on broad abilities like reasoning, memory in daily life or school and work performance. The one notable exception is narrow. In the ACTIVE trial, older adults who did an adaptive speed of processing exercise plus booster sessions had a lower rate of diagnosed dementia over 20 years, a result that has not been shown for a typical app routine.
If you enjoy the games, there is little harm. If you are paying to get smarter, protect your memory or ace a job test, the evidence does not back that promise. This guide reads the research the way we read AI benchmarks on trueiq: what was measured, against what control, and whether the gain shows up anywhere outside the test. We did not run our own training study; everything below comes from the regulators, trials and meta analyses linked in the text, all read on October 29, 2026.
The short answer: you get better at the games
Every serious review starts from the same finding: training works on the trained task. Play a memory game for six weeks and your score on it rises. The open question is transfer, whether that gain carries over to something you did not practice.
Researchers split transfer into two kinds:
- Near transfer: a gain on a different test of the same skill, such as a new working memory task after training on another working memory task.
- Far transfer: a gain on a different ability or on real life outcomes, such as reasoning, reading, grades, job performance or staying independent in old age.
Far transfer is what the marketing sells, and where the evidence runs thin. It is the same trap as an AI model tuned on a benchmark: the score rises, but the ability you cared about may not. Our guide to LLM benchmarks explained covers the machine version of this problem.
What the big reviews found
Simons and colleagues, 2016
The most thorough review is "Do Brain Training Programs Work?" by Daniel Simons and six coauthors in Psychological Science in the Public Interest (2016). They set out best practices for a training study and then checked every peer reviewed intervention study cited by the leading brain training companies listed on Cognitive Training Data, the site that hosted a 2014 open letter from 133 scientists and practitioners defending brain training. Their conclusion: extensive evidence that training improves the trained tasks, less evidence for closely related tasks, and little evidence for distantly related tasks or everyday cognitive performance. They also found that none of the cited studies met all of their best practices, and many had design or analysis flaws that rule out firm conclusions.
Sala and colleagues, 2019
Giovanni Sala, Fernand Gobet and coauthors ran a second order meta analysis in Collabra: Psychology, combining earlier meta analyses of working memory training, video games, music, chess and exergames across children, adults and older adults. Working memory training did produce near transfer. Far transfer effects were small or null, and once they accounted for placebo effects and publication bias, the far transfer effect was zero for every population and every kind of program.
A separate Sala team meta analysis focused on older adults (Intelligence, 2019) shows the pattern in numbers. It pooled working memory training studies with 2,140 participants:
| What was measured | Average effect size (Hedges' g) |
|---|---|
| The trained tasks | 0.877 (large) |
| Near transfer tasks | 0.274 (modest) |
| Far transfer tasks | 0.121 (near zero) |
| Far transfer, active control groups only | about 0 (reported as minus 0.008) |
When the comparison group also did something engaging, the far transfer gain disappeared.
Lampit and colleagues, 2014
A PLOS Medicine meta analysis by Amit Lampit, Harry Hallock and Michael Valenzuela pooled 52 randomized trials with 4,885 cognitively healthy older adults and found a small overall effect of computerized cognitive training on neuropsychological tests (g = 0.22). Effects showed up for processing speed, visuospatial skills, nonverbal memory and working memory, but not for executive function or attention. For app users, two design findings stand out: home based training was ineffective compared with group based training, and training more than three times a week was ineffective compared with three or fewer.
Lumosity: the FTC settlement and what it means
On January 5, 2016, the Federal Trade Commission announced that Lumos Labs, maker of Lumosity, would pay $2 million to settle charges of deceptive advertising. The FTC said the company claimed its games would improve performance at work, at school and in sports, delay age related cognitive decline and protect against dementia and Alzheimer's disease, and reduce impairment from conditions such as stroke, ADHD and PTSD, without the science to support those claims.
The order entered a $50 million judgment, suspended after the $2 million payment because of the company's financial condition. It requires competent and reliable scientific evidence before Lumos Labs makes future claims about real world performance, age related decline or health conditions, and it required notices to subscribers who signed up for auto renewal between 2009 and 2014. At the time, the FTC said subscriptions ran from $14.95 a month to $299.95 for a lifetime membership. We have not checked current Lumosity prices for this guide.
Lumosity's own homepage now describes a company funded study of 4,715 people (Lumosity five days a week for 10 weeks versus online crosswords) in which Lumosity users improved more on the company's assessments. The same page says more research is needed to connect better assessment scores to everyday tasks, and it flags experimenter bias, roughly 50% dropout in both groups and publication bias as open risks.
An independent trial points the same way. Kable and colleagues (Journal of Neuroscience, 2017) randomized 128 young adults to 10 weeks of Lumosity or ordinary online video games. The Lumosity group improved on the trained games, but both groups improved equally on standardized cognitive tests, about as much as people retested with no training at all. There was no effect on brain activity during decision making or on impulsive and risky choices.
BrainHQ and the ACTIVE trial: the strongest case
If any app has a real scientific claim, it is BrainHQ from Posit Science, and the claim rests on one study.
ACTIVE (Advanced Cognitive Training for Independent and Vital Elderly) enrolled 2,802 adults aged 65 and older in 1998 and 1999 and randomly assigned them to memory training, reasoning training, speed of processing training or a no contact control group. Training was up to 10 sessions of 60 to 75 minutes over five to six weeks, in small groups. Some participants were randomized to booster sessions at 11 and 35 months.
The results have been followed for two decades:
- 10 years (Edwards and colleagues, 2017): speed training was linked to a 29% lower risk of dementia compared with control (hazard ratio 0.71, 95% CI 0.50 to 0.998, P = .049). Memory and reasoning training showed no significant difference.
- 20 years (Coe and colleagues, February 2026): using Medicare claims for 2,021 participants, people in the speed arm who did at least one booster session had a lower risk of diagnosed Alzheimer's disease and related dementias (hazard ratio 0.75, 95% CI 0.59 to 0.95). Speed training without boosters showed no benefit (hazard ratio 1.01), and memory and reasoning training showed none either. According to the Johns Hopkins summary, 105 of 264 boosted speed trainees (40%) were diagnosed with dementia, against 239 of 491 controls (49%).
BrainHQ says the ACTIVE speed exercise was an earlier version of its Double Decision exercise, which is available only through a BrainHQ subscription.
Before reading that as "BrainHQ prevents dementia", note the limits:
- The outcome was diagnosis in insurance claims, not dementia confirmed by study clinicians.
- Boosters were offered only to people who finished at least eight of the first 10 sessions, so the boosted group was partly self selected, which the authors flag as possible bias.
- The control group had no contact at all. There was no placebo game, so part of any effect could come from attention, expectations or social contact.
- The training was supervised, in person and in groups. The National Academies' 2017 report, Preventing Cognitive Decline and Dementia, called the cognitive training evidence "encouraging although inconclusive" and said there was no evidence that the ACTIVE results apply to commercial computer based brain training apps, whose effects appear short term and limited to the practiced task.
The fair reading: one adaptive speed exercise, done in a structured way by older adults, has the best long term signal in the field. It is not evidence that a mix of daily brain games on your phone does the same.
Why gains rarely transfer
The 2014 consensus letter from the Stanford Center on Longevity and the Max Planck Institute for Human Development, signed by more than 70 scientists, gives the plain explanation. Getting better at a memory game often comes from learning a strategy for that particular game, which is rewarding but does not improve memory in general. A single task cannot stand in for a whole ability, which is a basic rule of psychological testing.
The Owen study in Nature (2010) showed this at scale. Of 52,617 BBC viewers who signed up, 11,430 completed a six week online study in which they were asked to train for at least 10 minutes a day, three times a week. Every trained task improved. Benchmark tests of reasoning and memory did not improve more than in a control group that answered obscure questions using online resources, and no transfer showed up even for untrained tasks closely related to the training.
The consensus letter also warns about the "file drawer effect": positive studies get published more often than null ones, so the published record likely looks better than reality. That is why the bias corrected Sala results matter.
Brain training vs cognitive assessment
| Item | Brain training app | Cognitive assessment |
|---|---|---|
| Goal | Practice tasks to raise your scores | Measure ability at one point in time |
| Changes with practice? | Yes, by design | Designed to resist practice, with parallel forms and norms |
| Validated against | Mostly its own games | Reference groups and clinical or job outcomes |
| Examples | Lumosity, BrainHQ | CANTAB from Cambridge Cognition, hiring aptitude tests |
A training app score tells you how good you have become at that app. A validated assessment like CANTAB compares you with a reference population, which is why drug trials use it. Hiring tests sit in between; our comparison of TestGorilla vs HackerRank covers what those tests measure and what research says about their validity. And if you are preparing for a coding screen, practicing the real task format beats general brain games, as our guide on how to take a coding test explains.
How to judge a brain training claim
Use the checklist the Stanford consensus and the Simons review suggest:
- Who ran the study? A single trial by researchers with a financial stake is not enough. Look for independent replication.
- What did the control group do? A no contact control flatters any program. An active control, such as video games or crosswords, is the fair test.
- What was the outcome? Better scores on the trained games prove little. Look for untrained tests and real life measures.
- How long did gains last? Most effects fade. ACTIVE needed boosters.
- Who was studied? Results in healthy adults over 65 do not automatically apply to students or workers in their 30s.
- What does it cost you in time? The consensus letter's advice is to count the opportunity cost: an hour of drills is an hour not spent exercising, learning a language or seeing friends. If drills replace television, the trade looks better. For general brain health, the letter says aerobic exercise has better support, with small but noticeable gains in cognitive performance.
FAQ
Does Lumosity work?
It makes you better at Lumosity games. Its own homepage says more research is needed to link better scores to everyday tasks, and an independent 2017 trial found no benefit over ordinary video games on standardized tests or decision making.
Is BrainHQ worth it?
BrainHQ has the strongest research link of any consumer app: its Double Decision exercise descends from the ACTIVE speed training linked to lower diagnosed dementia over 20 years in older adults who did booster sessions. That trial used supervised group sessions and a no contact control, so the result may not carry over to casual home use. Treat it as a structured speed exercise, not a proven dementia treatment.
Can brain training apps raise your IQ?
The evidence says no. The Sala second order meta analysis found that, after controlling for placebo effects and publication bias, far transfer effects on measures like reasoning were zero across all ages and program types.
What was the Lumosity FTC settlement?
In January 2016, Lumos Labs agreed to pay $2 million to settle FTC charges of deceptive advertising about real world and health benefits. A $50 million judgment was suspended, and the order requires competent and reliable scientific evidence for future claims about performance, cognitive decline or health conditions.
Can brain games prevent dementia?
No brain game has been shown to prevent dementia. The ACTIVE 20 year analysis linked one specific speed training regimen with boosters to a lower rate of diagnosed dementia, but the National Academies found no evidence that such results apply to commercial brain training apps.
How long should I train to see results?
On the trained games, scores improve within days to weeks. For anything broader, there is no proven dose. In the Lampit meta analysis, three or fewer sessions a week worked better than more, and group sessions worked better than training alone at home.
