Are puzzle games good for your brain? A little, in narrow ways, and less than the adverts say. Practising a puzzle makes you better at that puzzle. People who do a lot of word and number puzzles tend to score higher on thinking tests, but that is a correlation, and sharp people may simply be the ones who enjoy puzzles. A 2016 review of the brain-training research found that gains mostly stay with the task that was trained. None of the studies below shows that a daily sudoku prevents dementia.
I built PuzzleMoss, so I would like the answer to be yes. What follows sticks to what each study did and found, with links. It is not medical advice.
What "good for your brain" would have to mean
Researchers separate two kinds of improvement. Near transfer is getting better at the task you practised, or one very like it. Far transfer is the hope that practice spreads to something different, such as remembering names, staying independent in old age or a lower risk of dementia. Far transfer is what the marketing promises, and it is where the evidence thins out.
The second distinction is between observing people and assigning them. If puzzle fans score higher on memory tests, you cannot tell whether the puzzles helped or whether people with good memories stick with puzzles. Only a trial that randomly assigns people can start to pull those apart.
Puzzle fans do score better on tests
The biggest look at puzzle habits comes from PROTECT, an online study of adults over 50 run from the University of Exeter and King's College London. In two 2019 papers in the International Journal of Geriatric Psychiatry, Brooker and colleagues analysed 19,078 cognitively healthy volunteers aged 50 to 93, one paper on word puzzles and one on number puzzles. Participants said how often they did puzzles, on a scale from more than once a day to never, and took 14 online cognitive tests. More frequent puzzling went with better scores on all 14, and people who never did word puzzles scored lowest on every measure.
The university's press release described word-puzzle users performing like people ten years younger on some grammatical reasoning tests, the line the headlines used. The same release quotes the lead researcher, Anne Corbett, saying the study cannot show that puzzles reduce the risk of dementia. Puzzle frequency was self-reported, the tests were taken at one point in time, and the word-puzzle paper says the design cannot show whether puzzles change how thinking ages.
A Scottish study gives a clue about which way the arrow points. Staff and colleagues, writing in the BMJ in 2018, followed 498 people born in 1936 who had taken part in Scotland's 1947 Mental Survey as children, and tested them up to five times over about 15 years from late middle age. Those who reported more problem-solving activity, puzzles included, had higher scores, but they declined at the same rate as everyone else. Engagement was also linked to childhood ability and education. The authors read this as a higher starting point rather than protection from decline, and said plainly that an observational study cannot prove cause.
The newest large result is from Australia. Wu and colleagues (JAMA Network Open, 2023) followed 10,318 adults aged 70 and over from the ASPREE trial for about ten years. Those who often did "active mental activities", a group that covered games, cards, chess, crosswords and puzzles, had a 9% lower risk of dementia than those who did not, after adjusting for education, socioeconomic status and health. Adult literacy activities, such as taking classes or using a computer, were linked to an 11% lower risk. Social outings showed no association. These are modest differences, and people in the early, undiagnosed stages of decline may drop hobbies that start to feel hard, which would make the hobbies look protective.
What happens when people are assigned to train
In 2010 a team led by Adrian Owen ran a brain-training experiment with BBC viewers. Owen and colleagues (Nature, 2010) had 11,430 adults train online several times a week for six weeks on tasks aimed at reasoning, memory, planning, visuospatial skills and attention, alongside a control group that answered general-knowledge questions instead. Everyone got better at the tasks they practised. Scores on untrained benchmark tests did not follow, even where those tests were closely related to the training.
The most cited positive result is ACTIVE, a US trial that enrolled its participants in 1998 and 1999. Rebok and colleagues (Journal of the American Geriatrics Society, 2014) reported its ten-year follow-up. 2,832 independently living adults, average age 73.6, were randomly assigned to ten sessions of memory training, reasoning training, speed-of-processing training or no training, and some later had booster sessions. Ten years on, the reasoning and speed groups still did better on the abilities they had trained; the memory group did not. All three training groups reported less difficulty with everyday tasks than the control group, though that measure was self-reported.
A secondary analysis by Edwards and colleagues (2017) found that the speed-training group had a 29% lower risk of dementia over those ten years than the controls. A 20-year follow-up published in 2026, using Medicare claims records, linked speed training to fewer dementia diagnoses only in people who also had the booster sessions, and found no significant effect for memory or reasoning training. Dementia was never ACTIVE's primary outcome, the analyses have been disputed in print, and the 2026 authors say the result needs confirming with clinically assessed diagnoses.
Note what the speed training was: a computerised exercise in spotting things quickly on a screen, not sudoku or crosswords. Whatever ACTIVE shows does not carry over to puzzle games by association.
Crosswords against brain-training games
One trial has put a traditional puzzle head to head with computer games. Devanand and colleagues (NEJM Evidence, 2022) randomly assigned 107 older adults with mild cognitive impairment to either computerised crosswords or a set of web-based cognitive games, with 12 weeks of intensive training at home followed by booster sessions, and followed them for 78 weeks. On the main cognitive measure, a standard test called ADAS-Cog, the crossword group improved slightly and the games group worsened slightly. The gap was modest (P=0.04). The researchers had expected the games to come out ahead.
There was no group that did neither, so the trial can say crosswords beat these particular games, not that either beats ordinary life. It was small, and the authors call for a larger trial with an inactive control group. Everyone in it already had memory problems, so it says little about healthy adults.
Why the experts disagree
In October 2014 researchers convened by Stanford University and the Max Planck Institute for Human Development published a consensus statement signed by a long list of psychologists and neuroscientists. It said claims made for brain games were frequently exaggerated and at times misleading, and that while training improves the practised skills, it was not yet appropriate to conclude the gains go much beyond them. Another group of researchers, some with ties to the industry, issued a rebuttal.
Two years later Simons and colleagues reviewed the field in Psychological Science in the Public Interest, partly to work out how two groups of scientists could read the same studies and reach opposite conclusions. They found extensive evidence that training improves the trained tasks, less evidence for closely related tasks, and little evidence for distantly related tasks or everyday thinking.
Regulators came to a similar view. In January 2016 Lumos Labs, the company behind Lumosity, agreed to pay $2 million to settle US Federal Trade Commission charges that it had claimed, without the science to back it, that its games could improve performance at work and school and delay age-related cognitive decline. A $50 million judgement was suspended, and the order bars such claims without competent and reliable scientific evidence.
How to read the next headline
A few questions sort out most puzzle-and-brain stories:
- Were people randomly assigned to puzzle, or just asked about their habits?
- What was measured: the practised task, a lab test, or something in daily life such as a dementia diagnosis?
- How big was the difference, and compared with what?
- Who took part? Volunteers for online brain studies, or adults who already have memory problems, may not be like you.
Play them because you like them
None of this is an argument against puzzles. It is an argument against buying them as medicine. The trials do agree on one thing: what you practise is what improves. Get better at mahjong solitaire and you are better at mahjong solitaire, which is a perfectly good result.
So pick puzzles for the puzzle. If numbers and words are your thing, Brain & Numbers has sudoku, logic, maths and word puzzles. If you prefer looking closely, try hidden object and spot the difference. Mahjong and tile matching and block puzzles are there for pattern spotting and planning a move or two ahead. Our list of the best free puzzle games online is a reasonable place to start, and if mahjong is new to you, mahjong solitaire vs mahjong explains which one you are actually playing.
And if you, or someone close to you, is worried about memory, that is a conversation for a GP, not a puzzle website.