Do Puzzles Keep Your Brain Sharp? What the Research Says

Puzzles probably do not keep your brain sharp in the broad way the phrase suggests, but they are not useless either. People who do puzzles regularly tend to score better on thinking tests, yet those studies cannot show the puzzles caused it. Trials show you get better at what you practise, with little spread to other skills. No study shows puzzles prevent dementia. Enjoy them as one part of an active life.

This is the long version, for readers who want to see the evidence rather than take a headline on trust. It walks through the main studies people cite, what each one found, how strong it is and where it stops. It is written by the maker of a puzzle app, so it is worth saying plainly: nothing here is a reason to buy or play anything. If you only want a calm daily routine, the guide on building a daily puzzle habit is the practical companion to this page.

What "keeping your brain sharp" can mean

"Sharp" hides several different questions, and the evidence for each is very different. Before looking at any study, it helps to ask which one it answers.

Question	What a study would need to show	How strong the evidence is
Do puzzle players score better on thinking tests?	People who puzzle more score higher, at one point in time	Fairly consistent, but observational
Does doing puzzles make you better at puzzles?	Practice improves the trained task	Strong and unsurprising
Does puzzle practice improve other mental skills?	Gains on untrained tests (memory, reasoning)	Weak; small or no "transfer" in most trials
Does it help everyday life?	Better handling of money, medicines, driving and so on	Limited; some long-term trial signals, with caveats
Does it lower dementia risk?	Fewer diagnoses over years, ideally in a randomised trial	Not shown for puzzles; one trial signal for a specific speed-training task

Two words come up again and again. Observational studies watch what people already do and compare them. They are good at spotting links and bad at proving cause. Randomised controlled trials assign people to do one thing or another by chance, so differences afterwards are more likely to be caused by the activity. Most puzzle research is observational. Brain-training research includes many randomised trials, but they test computer tasks that are not quite the puzzles people play for fun.

The PROTECT study: puzzle players score better

The study most often quoted in news stories about sudoku and crosswords comes from PROTECT, a large online study of people aged 50 and over run by the University of Exeter and King's College London. In two papers in the International Journal of Geriatric Psychiatry, announced in May 2019, Helen Brooker, Keith Wesnes, Clive Ballard and colleagues asked more than 19,000 participants how often they did word puzzles and number puzzles, and gave them a set of online cognitive tests (University of Exeter, 2019).

The pattern was clear: the more often people said they did puzzles, the better they did on tests of attention, reasoning and memory, particularly on speed and accuracy. The researchers expressed the size of the difference in years. On grammatical reasoning, people who did word puzzles performed at a level typical of people about ten years younger; on short-term memory, about eight years younger. On problem-solving measures, people who regularly did puzzles performed on average like people eight years younger than those who did not (ScienceDaily, 2019).

Those numbers travelled widely, often as "puzzles make your brain ten years younger". The researchers themselves were more careful. Dr Anne Corbett of the University of Exeter Medical School, who led the research, said: "We can't say that playing these puzzles necessarily reduces the risk of dementia in later life" (University of Exeter, 2019).

What PROTECT can and cannot tell you:

It can tell you that, among tens of thousands of adults over 50, frequent puzzling goes together with better test scores. That is a real and consistent finding.
It cannot tell you which way the link runs. People whose thinking is already quicker may simply enjoy puzzles more and do them more often.
It cannot rule out other explanations. Volunteers for an online brain study, who answer questionnaires about their puzzle habits, may differ in education, health and motivation from people who do not.
The "years younger" figures compare average scores between groups. They are not a measure of how old anyone's brain is.

PROTECT follows participants over many years, so later papers may say more about change over time. For now, its puzzle findings are a snapshot.

Why a link is not proof: the jigsaw example

A neat pair of results from one study shows why observational links need caution. In 2018, Fissler and colleagues published a randomised trial in Frontiers in Aging Neuroscience with 100 cognitively healthy adults aged 50 and over. Fifty-two did jigsaw puzzles at home for at least an hour a day for 30 days, plus four sessions of cognitive health counselling; the other 48 had the counselling only (Fissler et al., 2018).

The trial found no difference: a month of daily jigsaws did not improve the group's overall visuospatial thinking compared with the control group (an effect size close to zero). Yet in the same participants, people's lifetime jigsaw experience was linked to better visuospatial thinking, even after allowing for age, education and other leisure activities.

Both findings are true at once. The lifetime link could mean that years of puzzling build something a month cannot. It could also mean that people with strong visual skills are drawn to jigsaws in the first place. The trial is the better test of cause, and over a month it found nothing.

Three things commonly produce a link without cause:

Reverse causation. Sharp thinking makes puzzles enjoyable, so sharper people puzzle more.
Confounding. A third factor, such as education, income, health or an active social life, makes people both more likely to puzzle and more likely to score well.
Early decline. Dementia begins changing the brain years before diagnosis. People in the earliest stages may quietly stop doing demanding hobbies, which makes "stopped doing puzzles" look like a cause when it is partly a symptom.

None of this means puzzles do nothing. It means a link alone cannot settle the question, which is why researchers turn to trials.

The ACTIVE trial: training works on what you train

One of the largest trials of mental exercise in older people is ACTIVE (Advanced Cognitive Training for Independent and Vital Elderly), run in the United States from 1998. It enrolled 2,832 people aged 65 and over, living in the community without major cognitive or physical decline, and randomly assigned them to one of three kinds of training or to a no-training control group: memory strategies, reasoning (spotting patterns in number series and solving problems), or speed of processing (a computer task of spotting objects quickly on a screen). Training was ten sessions over five to six weeks, and some people got "booster" sessions later (Ball et al., JAMA, 2002).

The first results, in JAMA in 2002, showed the pattern that runs through this whole field. Each type of training improved the ability it targeted: 87% of the speed group, 74% of the reasoning group and 26% of the memory group showed reliable improvement, and boosters added to the gains. But the gains stayed with the trained ability, and "no training effects on everyday functioning were detected at 2 years" (Ball et al., JAMA, 2002).

Later follow-ups are more interesting and more debated:

Follow-up	What was found	Limits
10 years (Rebok et al., 2014)	Reasoning and speed gains on their own tests were still measurable. All three groups reported less difficulty with everyday tasks than controls: about 60% of trained participants versus 50% of controls were at or above their starting level	Everyday function was self-reported; memory training no longer improved memory scores (Rebok et al., 2014)
20 years (Coe et al., 2026)	Linking the trial to Medicare records from 1999 to 2019, people who did speed training plus boosters had a 25% lower risk of a dementia diagnosis (hazard ratio 0.75). Memory and reasoning training, and speed training without boosters, showed no difference	Dementia was taken from Medicare billing claims, not clinical examination; people in Medicare Advantage plans were left out, leaving 2,021 people; the benefit was seen only in those who attended at least one booster (Coe et al., 2026)

The 20-year result, published in Alzheimer's & Dementia: Translational Research and Clinical Interventions in February 2026, stands out because it comes from a randomised trial rather than from comparing people's hobbies. It is still one result, from one specific computer task, in one subgroup. It says nothing directly about sudoku, crosswords or any game in a puzzle app, and the memory and reasoning training, which look more like puzzles, showed no effect on diagnoses.

Brain-training games and the transfer problem

If practice makes you better at the practised task, the big question is whether it spreads. Psychologists call this transfer: near transfer to closely related tasks, far transfer to different skills and everyday life. Commercial brain training rests on far transfer, and the evidence for it is thin.

Owen and colleagues, 2010. Working with the BBC programme Bang Goes the Theory, Adrian Owen's team had 11,430 adults aged 18 to 60 train online for six weeks, at least ten minutes a day, three times a week. People improved on every trained task, but the paper in Nature found "no evidence for transfer effects to untrained tasks, even when those tasks were cognitively closely related" (Owen et al., 2010).
Simons and colleagues, 2016. A review in Psychological Science in the Public Interest of 132 journal articles that brain-training proponents had cited in support of their claims, led by Daniel Simons at the University of Illinois, concluded that brain training improves the trained tasks, has less evidence for closely related tasks, and has "very little evidence" for distantly related tasks or everyday performance. Many studies had design flaws (Association for Psychological Science; University of Illinois).
The Lumosity settlement, 2016. In January 2016, Lumos Labs agreed to pay $2 million to settle US Federal Trade Commission charges that it had made unfounded claims, including that its games could delay age-related decline and protect against dementia and Alzheimer's disease. The FTC's Jessica Rich said the company "simply did not have the science to back up its ads" (Federal Trade Commission, 2016).
The Lancet Commission, 2024. Summarising a Cochrane review, the Commission found only low-quality evidence for small, immediate benefits of computerised cognitive training in healthy older people, and none for long-term benefits in maintaining cognition (Livingston et al., The Lancet, 202401296-0/fulltext)).

Logic puzzles are not brain-training software, but the same principle very likely applies. Solving many sudokus makes you better at sudoku: you scan faster and spot patterns such as pairs and X-Wings sooner, which is the whole point of a sudoku strategy ladder. Expect that skill to stay mostly with sudoku.

Crosswords versus brain games: one head-to-head trial

One trial compared a familiar puzzle directly with commercial brain games. D.P. Devanand and colleagues at Columbia and Duke randomly assigned 107 people with mild cognitive impairment (average age 71) to 12 weeks of home training on either web-based crossword puzzles or web-based cognitive games, followed by booster sessions up to 78 weeks. The results, published in NEJM Evidence in October 2022, favoured crosswords: better scores on the main cognitive measure at 12 and 78 weeks, better everyday function at 78 weeks and less brain shrinkage on MRI (Columbia Psychiatry, 2022).

The limits matter. There was no group that did nothing, so the trial shows crosswords did better than games, not that crosswords beat doing neither. The participants already had mild cognitive impairment, so the results may not apply to healthy adults. The authors themselves called for "replication in a larger controlled trial with an inactive control group."

An older observational study points the same way, with the usual caveats. In the Bronx Aging Study, Pillai and colleagues followed 488 people who started out without dementia. Among the 101 who later developed it, those who did crosswords at the start showed the onset of rapid memory decline about 2.54 years later than those who did not. The authors concluded that the role of crosswords "should be validated in future clinical trials" (Pillai et al., 2011).

What the Lancet Commission says about dementia risk

The best-known international overview of dementia prevention is the Lancet Commission on dementia prevention, intervention and care, led by Gill Livingston at University College London. Its 2024 report, published on 31 July 2024, lists 14 modifiable risk factors and estimates that together they account for nearly half of dementias, which in theory could be prevented by removing those risks (Livingston et al., The Lancet, 202401296-0/fulltext)).

The 14 factors are: less education, hearing loss, high blood pressure, smoking, obesity, depression, physical inactivity, diabetes, excessive alcohol, traumatic brain injury, air pollution, social isolation, and two added in 2024, untreated vision loss and high LDL cholesterol.

On mental activity, the Commission says it is "important to be cognitively, physically, and socially active" in midlife and late life, and its first recommendation pairs good-quality education for all with cognitively stimulating activity in midlife. The strongest evidence it cites is about education and work, not hobbies. In one analysis of 107,896 people, high cognitive stimulation at work was linked to a lower risk of dementia, and the Commission notes that stimulation at work "is of longer duration than cognitive interventions or cognitively stimulating hobbies" (Livingston et al., The Lancet, 202401296-0/fulltext)).

Two honest readings follow:

Puzzles fit the general advice. A regular puzzle habit is one way among many to stay cognitively active, and the Commission is clear that staying active matters.
Puzzles are a small part of a bigger list. Treating hearing loss, keeping blood pressure and cholesterol in check, staying physically active, not smoking and keeping up with friends all appear on the list as factors in their own right. A daily sudoku does not stand in for any of them. The Commission also cautions that some of the associations may be only partly causal.

What this means for your puzzle habit

Put together, the evidence supports a modest, cheerful view of puzzles. They are a good way to spend a few minutes, they make you better at puzzles, and they may be part of an active, engaged life that goes with better thinking in later years. They are not medicine.

Claim you may see	What the evidence supports
"Puzzles make your brain ten years younger"	Frequent puzzlers in PROTECT scored like people several years younger on some tests; cause is not shown
"Sudoku prevents dementia"	No study shows this
"Brain games improve your memory and work performance"	Gains mostly stay with the trained task; Lumosity settled with the FTC over claims like this
"Crosswords beat brain games"	One trial in people with mild cognitive impairment found this, with no do-nothing group
"Staying mentally active is good for you"	Supported as general advice by the Lancet Commission, alongside physical and social activity
"Any brain training lowers dementia risk"	One trial signal for one speed-training task with boosters; memory and reasoning training showed none

If you want your puzzle time to line up with what the research does support, a few habits make sense:

Keep it enjoyable. Enjoyment is what keeps a habit going for years, and any long-term benefit, if there is one, would take years.
Move up the levels. Staying with puzzles that still make you think is closer to "cognitively stimulating" than repeating ones you can do on autopilot. The guide to puzzle difficulty levels explains what changes from Easy to Expert in each game.
Vary the puzzle. Different puzzles use different kinds of reasoning: digits and elimination in sudoku, spatial planning in a path puzzle, counting and geometry in Bridges. If you are curious where to start, see the best logic puzzles for beginners.
Make it social. Social isolation is on the Lancet list; comparing the same puzzle with a friend turns a solo hobby into a small shared ritual. The guide to puzzles with friends has ideas.
Keep the rest of the list in view. A walk, a hearing test or a phone call with a friend each address a factor that is on the Lancet list in its own right.

Frequently asked questions

Does doing sudoku prevent dementia?

No study has shown that sudoku or any other puzzle prevents dementia. The PROTECT study found that people over 50 who did number puzzles more often scored better on thinking tests, but the researchers said they could not conclude that puzzles reduce dementia risk. The Lancet Commission recommends staying cognitively active as general advice, without singling out puzzles.

Are crosswords better for the brain than number puzzles?

There is no good head-to-head evidence between crosswords and number puzzles. PROTECT found links for both, on slightly different tests. The one trial that favoured crosswords compared them with computer brain games, in people who already had mild cognitive impairment, and had no do-nothing group.

Do brain-training apps work?

They make you better at the games they contain. Large studies and reviews, including Owen and colleagues' 11,430-person trial and the 2016 review by Simons and colleagues, found little evidence that those gains spread to untrained tasks or everyday life. In 2016 the US Federal Trade Commission settled with Lumosity over claims that its games could delay cognitive decline.

How often should I do puzzles to keep my mind active?

No study gives a dose. In PROTECT, more frequent puzzling went with better scores, but that does not mean more puzzles would raise yours. A practical answer is: often enough that it is a regular habit you enjoy, and challenging enough that you still have to think.

Is it too late to start puzzles in my 60s or 70s?

It is never too late to enjoy puzzles or to get better at them; the ACTIVE trial participants were 65 and over and improved on the skills they trained. Whether starting late changes long-term risk is not known. The Lancet Commission's advice to stay cognitively, physically and socially active covers later life as well as midlife.

Where to go next

The honest summary fits on one line: puzzles are good for puzzling, possibly part of a mentally active life that goes with better ageing, and not a treatment for anything. That is reason enough to enjoy them.

If you want to turn that into a routine, start with how to build a daily puzzle habit. If you want to understand why a puzzle can feel restful even when it is hard, read why logic puzzles feel calming. And if a hard grid is what keeps you thinking, the sudoku strategy guide takes you from singles to X-Wing. Puzzle Picnic's six puzzles are free to play in the browser at puzzlepicnic.app, with four levels each.