Do Brain Training Apps Work? The FTC Has Opinions

Brain training reliably makes you better at the exercises you practice. What it has not been shown to do is transfer — the largest single study, 11,430 adults training online over six weeks, found “no evidence… for transfer effects to untrained tasks, even when those tasks were cognitively closely related.”¹ The most thorough review of the field concluded there is “extensive evidence that brain-training interventions improve performance on the trained tasks, less evidence that such interventions improve performance on closely related tasks, and little evidence that training enhances performance on distantly related tasks or that training improves everyday cognitive performance.”³ In January 2016 the FTC charged Lumosity with deceptive advertising over exactly this gap, and the company settled.⁵

Written by the Fog & Clarity editorial team · Published 13 August 2026 · Last updated 13 August 2026

How it’s supposed to work

The premise is straightforward and not unreasonable. Cognitive abilities like working memory, processing speed and reasoning are treated as capacities — something like muscles. Exercise the capacity with a task that loads it hard, and the capacity itself grows. Because that capacity is used by everything else you do, the benefit should show up in unrelated places: remembering where you parked, following a meeting, doing your job.

Researchers call the step from the trained game to everything else transfer, and they grade it by distance:

Practice effects — you get better at the exact game you played. Universally observed, and not in dispute.
Near transfer — improvement on a different task measuring the same narrow ability. Sometimes observed.
Far transfer — improvement on a distantly related ability, like fluid intelligence. Mostly not observed.
Everyday transfer — improvement in real life: work, school, driving, managing a household. The thing anybody actually wants, and the weakest evidence of the four.

The entire commercial claim rests on the last two rungs. The entire scientific literature is strongest on the first.

What the FTC actually did

On 5 January 2016 the Federal Trade Commission announced a settlement with Lumos Labs, the maker of Lumosity. The FTC charged that the company had falsely claimed its games would improve performance “on everyday tasks, in school, at work, and in athletics,” and would “delay age-related cognitive decline and protect against mild cognitive impairment, dementia, and Alzheimer’s disease,” along with reducing impairment from stroke, traumatic brain injury, PTSD, ADHD, chemotherapy side effects and Turner syndrome.⁵

The stipulated order carried a $50 million judgment, suspended to a $2 million payment on the basis of the company’s financial condition. It required the defendants to hold “competent and reliable scientific evidence before making future claims about any benefits for real-world performance, age-related decline, or other health conditions,” and to notify auto-renewal subscribers from 2009 through 2014 of the action and give them a way to cancel. Memberships had run from $14.95 a month to $299.95 for a lifetime subscription.⁵⁶

Four months later, on 18 May 2016, the FTC settled with LearningRx Franchise Corp. over one-on-one brain training programs. The Commission alleged the company had claimed its programs permanently improved serious conditions including “ADHD, autism, dementia, Alzheimer’s disease, strokes, and concussions,” and substantially raised school grades, test scores, earnings and athletic performance. The order suspended a $4 million judgment on payment of $200,000, and prohibits cognitive-improvement claims “unless the claims are not misleading and substantiated by human clinical testing.”⁸

Two things are worth being precise about. Neither order says brain training does nothing — a settlement is not a scientific finding, and the FTC’s authority is over advertising claims, not over games. What both orders say is that the claims being made ran well ahead of the evidence supporting them. That is a narrower statement and a more durable one. It is also the same pattern that produced the December 2024 court order in the Prevagen case, in a different corner of the same market.

What the evidence says

The field is unusually well studied for a consumer product, which is what makes the gap so legible.

Study	Design	Size	Finding
Owen et al., Nature 2010	6-week online training, 4 benchmark tests	11,430 adults, 18–60	Every trained task improved; no transfer to untrained tasks
Melby-Lervåg et al., 2016	Meta-analysis of working-memory training	87 publications, 145 comparisons	Near transfer yes; no convincing far transfer vs treated controls
Simons et al., 2016	Systematic review of the whole field	Field-wide	Little evidence of everyday cognitive benefit; major design flaws common
Cochrane (Gates et al., 2020)	Computerised training ≥12 weeks, healthy 65+	8 RCTs, 1,183 participants	Small short-term global-cognition benefit; low to very low certainty
ACTIVE 10-year (Rebok et al., 2014)	10 sessions plus boosters, 10-year follow-up	2,832 adults, mean age 73.6	Reasoning and speed maintained; memory not; small IADL difference

The study that set the tone

Owen and colleagues recruited through a BBC television programme and ended up with 11,430 healthy adults aged 18 to 60 who completed six weeks of online training, a minimum of ten minutes a day, three times a week — an average of about 24 sessions each. Four benchmark tests covering reasoning, verbal short-term memory, spatial working memory and paired-associates learning were administered before and after.¹

Improvements appeared in every trained task. They did not appear anywhere else. The authors’ summary of the practical magnitude of what did change is the sentence the field has never quite escaped: on the observed rate of gain, “it would take almost four years of training to remember one extra digit.”¹

The meta-analyses agreed

Melby-Lervåg, Redick and Hulme pooled 87 publications and 145 experimental comparisons of working-memory training. They found “reliable improvements on measures of intermediate transfer (verbal and visuospatial working memory)” — the near rung — and, on measures of real-world cognitive skills, “no convincing evidence of any reliable improvements when working memory training was compared with a treated control condition.” Their conclusion was that such programs “produce short-term, specific training effects that do not generalize to measures of ‘real-world’ cognitive skills.”²

The phrase “compared with a treated control condition” is carrying weight there. Comparing trainees against a group that did nothing at all confounds training with attention, novelty, computer familiarity and the expectation of improvement. Simons and colleagues, reviewing the field in Psychological Science in the Public Interest, found that “many of the published intervention studies had major shortcomings in design or analysis” and that “none of the cited studies conformed to all of the best practices.”³ Where an active control group is used, the effects tend to shrink.

What Cochrane found in older adults

The 2020 Cochrane review looked specifically at computerised cognitive training run for 12 weeks or longer in cognitively healthy people aged 65 and over: 8 randomised trials, 1,183 participants, interventions of 12 to 26 weeks. Small benefits to global cognitive function were possible immediately after training compared with active controls, and a slight episodic memory improvement versus no activity, “of uncertain clinical importance.” There was no evidence that anything persisted 12 months after training ended. The review authors rated the certainty of the evidence “low or very low for all outcomes.”⁴

The exception worth knowing about

One trial genuinely stands apart, and honesty requires giving it its due. ACTIVE randomised 2,832 community-dwelling older adults, mean age 73.6, to memory training, reasoning training, speed-of-processing training, or a no-contact control. Training was 10 sessions over five to six weeks, with four booster sessions at 11 months and four more at 35 months. Participants were followed for ten years.⁹

At ten years, the targeted abilities held up for two of the three arms: speed of processing at an effect size of 0.66 and reasoning at 0.23, with memory training at 0.06 and not maintained. Each of the three interventions was associated with less self-reported decline in instrumental activities of daily living than the control group, effect sizes 0.36 to 0.48 — roughly 60% of trained participants at or above their baseline IADL function at year 10 against 50% of controls.⁹

Three qualifications belong with those numbers, and they are why ACTIVE does not rescue the consumer category. The outcome that carried into daily life was self-reported, and self-report from unblinded participants who know they were trained is a soft measure. The control group received no contact at all rather than an active alternative. And the trained abilities were narrow, specific ones taught in supervised sessions by instructors — not an app subscription with a daily streak counter. Whatever ACTIVE demonstrates, it is not that the thing sold in an app store has a ten-year benefit.

How to actually do it, if you’re going to

The evidence is not “training is useless.” It is “training improves what you train.” That is a usable principle if you invert the usual approach: stop trying to upgrade a general capacity and train the specific thing you want to be better at.

Name the actual failure. Not “my memory is bad” but “I lose people’s names within a minute of being introduced,” or “I lose the thread in long meetings.” The specific version is trainable; the general version is what apps sell against.
Train that, not a proxy for it. Names: practice retrieving names, spaced out, with real faces. Meetings: practice structured note-taking and summarising immediately afterwards. A dual n-back task is a proxy, and proxies are precisely what fail to transfer.
Use spacing and retrieval, not exposure. The reliable finding in learning research is that testing yourself with gaps between attempts beats re-reading or passive practice. Any tool that makes you retrieve is doing the work; the graphics are not.
Give it a real dose. ACTIVE’s protocol was ten sessions over five to six weeks with boosters nearly a year later. That is the shape of the only long-term positive result in the field: short, structured, and revisited — not fifteen minutes a day forever.
Measure the real outcome, not the in-app score. Your score in the game will rise. That is the practice effect and it tells you nothing. Count the thing you actually cared about in step one, over a month.
Check the free version first. Nothing in the evidence base distinguishes a paid program from an unpaid one. If a subscription is the difference between doing it and not doing it, that is a motivation purchase, and it is fine to call it that.

A worked example. Someone who keeps forgetting names spends four weeks on a spaced-repetition deck of colleagues’ photographs, reviewing for five minutes daily, and keeps a tally of introductions where the name was still available an hour later. Week one: 2 of 9. Week four: 6 of 10. That is a real, measured, near-transfer gain in the exact skill that was failing — and it is also the sort of result no brain-training study has ever needed to claim, because it never left the trained domain in the first place.

Where it fails

As a substitute for a cause you haven’t identified. Most complaints of fog and forgetfulness in adults under 60 are not a training deficit. Sleep debt, circadian misalignment, medication, alcohol, untreated hearing loss, depression and thyroid problems all present this way, and none improves because you played a matching game. If your symptoms are new, worsening, or accompanied by anything on the list at brain fog red flags, that is a doctor’s question, not an app’s.

As reassurance. A rising score is designed to feel like progress, and it is the one thing every study agrees is not evidence of anything beyond itself. That feedback loop is the product.

As prevention. No claim of the kind the FTC charged Lumosity with — delaying age-related decline, protecting against dementia — is supported by the evidence, and both the 2016 order and the 2020 Cochrane review are clear about the state of it.

As displacement. The realistic cost of an app is not the subscription; it is the twenty minutes and the belief that the problem is being handled. If those twenty minutes came out of sleep, they made things worse. Sleep debt and the afternoon dip are the far more common explanations for the symptom the app is being bought to fix.

Why there is nothing to buy on this page

This site earns money from supplement affiliate links, and it says so on the pages where those links appear. This is not one of them, and neither is anything else in our methods section. That is deliberate. There is no product here we could recommend on the evidence, and a page arguing that brain training doesn’t transfer would be worth nothing if it ended by selling you something instead. The parts of this site that make no money are the reason to trust the parts that do.

What would change our mind

A large, preregistered, adequately powered randomised trial with an active control group — matched for time, attention, novelty and expectation — showing a durable improvement on an objectively measured everyday outcome rather than a self-reported one, and holding up at follow-up months after training stopped. That trial would settle the question in one direction, and the field has been asking for it since at least 2016. If it is run and it comes out positive, this page gets rewritten.

Frequently asked questions

Is Lumosity a scam?

That is not a claim we can substantiate and not what the record says. What the record says is that in January 2016 the FTC charged Lumos Labs with making advertising claims about real-world performance and cognitive decline that were not supported by competent and reliable scientific evidence, and that the company settled under a $50 million judgment suspended to a $2 million payment. The games work as games. The claims made for them were the problem.

Are crosswords and sudoku any different?

They face the same transfer problem: you get better at crosswords. The honest difference is that nobody sells a crossword as dementia prevention, and doing something you enjoy that also involves other people has separate reasons to recommend it. Treat it as a pastime with a possible side benefit, not an intervention.

What about learning a language or an instrument?

Also unproven as a general cognitive intervention, and also a much better use of the same hours — because the near-transfer gain is the point. You end up able to speak Portuguese, which is a real skill rather than a proxy for one. Judge it as learning, not as training.

Does brain training help children with ADHD?

This page covers healthy adults and does not extend to any clinical population, and the FTC has taken action against exactly this claim being made without human clinical testing behind it. If ADHD is the question, it is one for a clinician who can assess the individual, not one an app or an article should answer.

Would a supplement work better?

No. The evidence for cognitive supplements in healthy adults is, if anything, thinner — not one compound in our reference file reaches a strong rating for improving cognition in healthy people. Our tier list and the case against brain supplements set out where that stands, and we have a commercial interest in those pages that we do not have in this one.

Sources

These statements have not been evaluated by the Food and Drug Administration. Nothing here is intended to diagnose, treat, cure or prevent any disease.

This article is information, not medical advice. Talk to a doctor or pharmacist about your own situation, especially if you take prescription medication.